Volume 1 No 1

2007 Journal of Computer Science Vol 1, No 1, June 2007, pp 05-10 Faculty of Computer Science ISSN 1994-6244
Capacity and Coverage Calculation Model for the UMTS

M. Shamim Kaiser1, Md. Ezharul Islam2 and Md. Raihan Jamil1
ABSTRACT
This paper deals with the coverage and capacity issues of upcoming 3G (Third Generation) or UMTS (Universal Mobile Communication Telecommunication System) and discuss the hurdles to implement the UMTS or 3G system in the place of existing 2G radio network. The radio interface of UMTS operating in the WCDMA (Wideband Code Division Multiple Access) FDD (Frequency Division Duplex) mode required the updating of 2G radio network. In this work an attempt has been made to find the total number of users that acquired the maximum adequate signal strength in a cell. Finally Optimization of coverage and capacity calculations for the UMTS has also been done. Keywords: UMTS, Network Planning, Capacity and Coverage, Propagation Model and WCDMA
I. INTRODUCTION 3G radio network, is also referred to as UMTS, is adopted to fulfill the user requirement for innovative services such as enhanced and multimedia messaging through high-speed data channels [1]. Wide-band CDMA (WCDMA), CDMA2000 and Time-Division Synchronous CDMA (TD-SCDMA) are the three standards accepted by ITU-T (International Telecommunication Union, Telecommunication Standard). Owing to the high costs and the scarcity of radio resources, an accurate and efficient mobile network planning procedure is required. The objective of network planning is to maximize the coverage, capacity and the quality of service (QoS) [2]. The planning of UMTS (or 3G) is very different from the 2G which adopts a time division multiple access (TDMA) technology. TDMA works by dividing a radio frequency into time slots and then allocating slots to multiple calls. In this way, a single frequency can support multiple, simultaneous data channels. 2G planning can be divided into a coverage planning phase and a frequency planning phase which are driven by coverage and capacity criteria respectively [3]. In the coverage planning phase, base stations (BSs) are placed so that the signal strength is high enough in the area to be served. In the frequency-planning phase, each BS is assigned a number of channels, and then network operator takes into account the traffic load, the level of services measures as the signal-to-interference ratio (SIR) as well. However, in the 3G planning, since all carriers in the network use the same frequency range, frequency planning is not required. Furthermore, coverage and capacity planning should be performed in tandem since capacity requirement and traffic distribution influence the coverage. UMTS or 3G mobile system has a significant impact not only on the RF (radio frequency) network, but also on the core network architecture. Care must be taken to allow current GSM (global system mobile) operators to protect their infrastructure investments when their networks are upgraded to support UMTS. This paper focuses on the differences between GSM radio system planning and UMTS radio system planning. UMTS uses WCDMA (wideband code division multiple access) as the radio transmission technology. It is claimed that TDMA (time division multiple access) RF planning is much more difficult than WCDMA-based systems. This is true, in part, because of the interference issues. However, UMTS serves users with various demands, and many aspects of planning are more closely interrelated to each other in UMTS planning than they are in GSM planning [1-5]. The major differences in the UMTS radio system planning process occur in coverage and capacity planning. In GSM, coverage is planned separately after the network is dimensioned (based on the market study), and capacity and frequency are planned in tandem. In UMTS, coverage and capacity are planned at the same time, because capacity requirements and traffic distribution influence coverage. Frequency and code can be planned separately. On the other hand, the wideband nature of WCDMA technology (5 MHz) compared with GSM (200 kHz) imposes new criteria in modeling the propagation environments. This paper outlines the challenges and solutions for planning and optimizing UMTS networks with respect to radio interface. WCDMA air interface specifications
1. 2. Telecommunication Engineering. Training and Instrumentation Division, University Grants Commission of Bangladesh, Agargaon, Dhaka-1207, Bangladesh. e-mail: shamimkaiser@gmail.com Computer Science and Engineering. Training and Instrumentation Division, University Grants Commission of Bangladesh, Agargaon, Dhaka-1207, Bangladesh. e-mail: mdezaharul@yahoo.com
Md. Amiruzzaman et al.
and their implications on transmission channel behavior and modeling are described first. Next, solutions are provided for system design, including coverage, capacity, code, and frequency planning. The analysis captures both design processes and engineering calculations. The critical optimization and monitoring of WCDMA network performance are then discussed. II. UMTS ARCHITECTURE The UMTS network architecture is depicted in Figure 1. The core network handles call control and mobility management functionalities, while the UTRAN (UMTS terrestrial radio access network) manages the radio packet transmission and resource management. Packet routing and transfer within the core network are supported by definition of new logical network nodes called GGSN (gateway GPRS [general packet radio system] support node) and SGSN (serving GPRS support node). The GGSN is basically a packet router with additional mobility management features, and it connects with various network elements through standardized interfaces. The GGSN acts as a physical interface to the external packet data networks. The SGSN handles packet delivery to and from mobile terminals. Each SGSN is responsible for delivering packets to the terminals within its service area. GGSN and SGSN are capable of supporting terminal data rates up to 2 Mbps. A UTRAN consists of one or more RNSs (radio network subsystems), which in turn consist of base stations (Node Bs) and RNCs (radio network controllers). The RNS performs all of the radio resource and air interface management functionalities. The UMTS network architecture inherits most of its structure from the GSM model in the UTRAN.
Figure 1: UMTS Architecture [4-6] III. UMTS RADIO NETWORK PLANNING The UMTS radio network planning process can be divided into three parts; the initial phase (also called system dimension); detailed planning phase; and the optimization and monitoring phase. Each of the phases requires additional considerations such as propagation measurements, traffic demand measurements and so on. The process of the UMTS radio network planning is illustrated in Figure 2.
It is observed that the whole process needs to take into account coverage and capacity planning. In a cellular system where the entire air interface connections operate on the same carrier, the number of simultaneous users is directly influences the total noise of receivers. It is well known that the coverage of the cell has an inverse relationship with the user capacity of the same cell. An increase in the number of active users in the cell causes the total interference seen at the receiver to increase. This causes an increase in the power required to be received from each user. This is due to the fact that each user has to maintain a certain SIR at the receiver for satisfactory performance. For a given maximum allowable transmission power, an increase in the required power reception will result in a decrease in the maximum distance a mobile can be from the base station thereby reducing the coverage. The up-link and the ideal situation of the WCDMA technology are considered. That is, It is assumed that the coverage is limited by the maximum transmit power at the mobile and no blocking and outage take place in the cellular system since WCDMA technology can provide the enough codes for the new call to the cell in the ideal situation.
12
JournalofComputerScience,Volume1,Number1,June2007,ISSN:19946244
Embedded Wireless Network-Engine for VOD Sharing between Resource Constrained Devices
A. Initial Model for Capacity Calculation In order to calculate the maximum number of users in a particular cell the following assumptions are made: 1. No inter-cell and intra-cell interference is present with in the cell. 2. All signals arrive at the base station with equal power. 3. Un-limited number of spreading codes are available. The actual capacity of a WCDMA cell depends on many different factors, such as power control accuracy, interference power. If there are N users in a cell and the signal is denoted by S then the interference can be calculated as I = ( N 1) S + , where is the thermal noise. Hence the SIR is given by
SIR =
S 1 (1) = ( N 1 )S + ( N 1 ) + / S
The value of the interference is based upon the three assumptions mentioned above and is not realistic. The inter-cell interference is attributed to mobile stations and base stations in one cell affecting the operations in another cell. Intra-cell interference is attributed to the effect mobile stations and has the transmitted signal of other mobile stations within the same cell. It is a simple step to change the interference value to consider the other two factors though in this paper it would only serve to complicate the formulae. Suppose the digital demodulator for each user can operate against the noise at energy per bit-to noise power density level is given by E b / I o , where Here E b = S / R and I o = I / W (2) where, W is the chip rate and R is rate of data communication. Hence using equations (1) and (2) gives
(N 1) =
W/R Eb / I o S
(3)
In particular, if the user is not speaking during part of the conversation, the output of the coder is lowered to prevent the power from being transmitted unnecessarily. For a uniform population, this reduces the average signal power of all users and consequently the interference received by each user. The capacity is then increased proportional to this overall rate reduction. Speech statistics shows that a user in a conversation typically speaks close to 38% of the time using DTX (discontinuous transmission) and DRX (discontinuous reception) method controlled by transcoder (TC) as it has VAD (voice activity detection). The effects of voice activity and sectorization should be considered. This results in an increase in the Eb / I o by a voice activity gain factor, G v . Similarly, the sectionzation gain factor G s also increases the Eb / I o and the equation for the effective number of users, N , becomes
(N
W /R 1) = Gv G s E / I S b o
(4)
Using speech statistics we can set G v =2.63 [1] and based upon the usual three sectors for a cell,
G s =2.63.
W/R is Eb / I o
typically a value between 20 to 100.. Finally, for a cellular system in which all users in all cells employ the common spectral allocation of W, we must evaluate the interference. Normally the total interference equal the amount of interference from other cells and interference from given cells. Mathematically,
Due to the interference, the actual numbers of user will decrease from ( N 1 ) to ( N 1 ) [7], the equation (4) will be modified into
W /R 1 ( N 1) = Gv G s E / I S 1+ f b o
(5)
Normally, the total interference from users in all other cells equals approximately three-fifths of that caused by all users in the

13
given cell [9]. That is to say,
1 + f 1.6 . However we note that equation (5) gives a poor estimate to N due to the
many assumptions in the model. It does not reflect the relationship between the capacity and the coverage. B. Coverage versus Capacity The analysis in the above the capacity calculation can be isolated from coverage. However, in the 3G networks, it is important to derive the relationship between the coverage and capacity. The up-link and the ideal situation of the WCDMA technology are considered. That is to say, we will assume the coverage is limited by the maximum transmission power at the mobile and neither blocking nor outage takes place in the cellular system. Based on Equation (5), we can understand the performance of a WCDMA network by developing a simple expression for the ratio E b / I o as follows.
Eb S/R = I o ( N 1 )S ( 1 + f )Gv G s + W
Solving the equation (6) for S gives
(6)
S=
Eb R Io E W ( b R( N 1 )( 1 + f ) / G v G s Io
(7)
This equation can be used to demonstrate the dependency of S on the user data rates, R the voice and section gain G v , G s , respectively, and the total number of active users in a cell, N . We focus on the coverage by user 1 when the number of users in the cell is N . Let r be the distance of user 1 from the base station. The received power at the base station from mobile user l, S, is given by
S = S 1 P( d ) Z (8) where, S 1 The transmission power of the user, P( d ) is the propagation loss at distance d from the MS to BS, Z The
shadow fading The propagation loss is usually calculated by using propagation models. The propagation model that is most commonly used is the Okumura-Hata. The model is developed by combining propagation theory and extensive measurement campaigns. The model considers several parameters such as effective antenna height, terrain type, terrain height, frequency, and so on. For details of the Okumura-Hata model the reader is referred to [2, 3]. The model predicts
(9) Where A, B are constants defined by the frequency and
f is the frequency. a( hms ) is a function that depends upon the height of the mobile station that is particular to the environment. C m is the area type correction factor. Equation (9) shows
the relationship between the propagation loss and the distance and can be expressed as
P( d ) = p 1 + p 2 log 10 d
(9)
Based on equations (7), (8), (9) the relationship among the received power, number of users, the coverage area, the gain of voice and gain of sectionzation is given by=
Eb R Io E W ( b R( N 1 )( 1 + f ) / G v G s Io
(10)
14
IV. UMTS (WCDMA) TRANSMITTER, RECEIVER, AND CHANNEL PARAMETERS Link budget is the main tool for planning WCDMA coverage. WCDMA link budget, takes into account the base station equipment and the base station antenna line configuration. A typical link budget for WCDMA is presented in Table 1. The link budget in Table 1 is divided into five parts. In general information, the frequency band, chip rate, and temperature constant are given. In service information, the bit rates and loads for uplink and downlink are defined. Receiving end and transmitting end define the radio links in the uplink and downlink directions, respectively. Finally, isotropic path loss is defined as the maximum expected path loss between the receiver and transmitter [6]. WCDMA Coverage and Capacity Planning Coverage and capacity planning in WCDMA are interrelated. In low traffic areas, WCDMA planning is quite similar to GSM planning, because the load does not have a great impact on coverage. Of course, many details differ between the systems, but the main principles can be applied to both. In high traffic areas, unlike for GSM, there is no clear split between coverage, interference, and capacity planning of WCDMA. A. Coverage Planning The propagation predictions for WCDMA require the same planning phases as in GSM. First, the base station configuration and the link budget have to be defined. Also, the coverage threshold has to be well defined to exceed the required quality criteria but avoid unnecessary additional investments for the radio network elements. Moreover, the capacity targets and forecasts have to be well known at this phase because they have a strong effect on the base station coverage area. When the base station antenna height, coverage threshold, and capacity requirements are defined and the base station configuration is clarified in the link budget calculations, the actual propagation predictions process can start. Propagation measurements can be performed to fine tune the propagation prediction model. When the prediction model is tuned, the final base station parameters can be used to make the propagation predictions. Optimized base station parameters can be evaluated when the planning criteria are defined. This planning threshold means that agreement must be reached on the reasonable QoS level required for the different geographical locations. The threshold also depends on whether the service has to be extended inside vehicles and buildings in different areas. The planning threshold is defined in GSM by starting from the mobile station sensitivity and by adding the required clutter planning margins to the sensitivity value in each particular planning terrain bin. B. Capacity Planning WCDMA capacity planning is directly related to the link budget and, thus, to the base station coverage area. In the link budget in Table 1, only one type of service (64/144 kbps data transmission) was introduced, and the base station coverage was fixed for this service. It is possible to have any type of service between the voice calls and 2 Mbps data traffic in the WCDMA base station. This means that the base station coverage area is different for different users. (Figure 3) General Value information Frequency 2100MHz Chip rate 3.84Mcpc Temperature 0 oC Service Urban information Uplink Downlink Load 30% 50% Bit rate 64Kbps 144Kbps Transmitting end Tx power/ connection Tx power Tx antenna gain Peak EIRP Isotropic path loss
Receiving end Thermal noise density Rx noise figure Rx noise density

Urban Uplink Downlink 0.126W 1W 21dBm 30dBm 0 18dB 21dBm 44dBm 154.32dB 156.34dB
Urban Downlink 173.93dBm/Hz 6dB 167.93dBm/Hz
Uplink 173.93dBm/Hz 3dB 170.93dBm/Hz
15
Noise power Interference margin Receiving end Receiver interference power Processing gain Required Eb/No Receiver sensitivity Cable Loss Rx antenna gain Soft handoff diversity gain Required signal power Field strength
-105.09dBm 1.55dB Urban Uplink -108.54dBm 17.78dB 5dB -116.32dBm 4dBm 18dB 3dB -133.32dBm 10.32dB V/m
-102.09dBm 3.01dB Downlink -102.09dBm 14.26dB 4dB -109.34 dBm 0 0 3dB -122.34dBm 31.31dB V/m
Table1: Typical link budget for UMTS cell
Figure 3: Relative cell range and cell area versus user bit rate using WCDMA (cell ranges calculated by Okumura-Hata model at antenna height of 25m) Basically, the question is about the spreading power factor, SPF, which varies significantly when comparing the 12.2 kbps voice call (SPF = 25 dB) and 2 Mbps data transmission (PG = 2.8 dB) connections. In the uplink direction, the main objective in capacity planning is to limit interference from the other cells to an acceptable level. Network planning can increase the uplink load by reducing other cell interference. This can be achieved by obstacles to block the interfering cells. Also, downtilting is a very useful tool in limiting interference. In the downlink direction, two aspects should be considered: the interference from other cells and the power of the base station. The load equation for the downlink is similar to the equation for the uplink. However, in the downlink there is a new parameter called orthogonality and users are much more orthogonal compared with uplink. V. CONCLUSION AND FUTURE WORKS The efficient network planning is a vital aspect of UMTS networks. 2G and 3G networks are different from each other owing to the different levels of service offered. This article includes the optimization of capacity, coverage and quality of service. The equations derived in this work can be used in cellular system planning to set hard limits on the maximum number of users that can be admitted into the cell. Coverage and capacity in WCDMA are heavily coupled and can not be planned separately. The relationship given in (10) describing the relationship between coverage, capacity and data rates is very useful in planning the networks. However, there are shortcomings in the analysis and modeling provided herein. The modeling is largely based upon an ideal situation, wherein no call will be blocked. While WCDMA technology can provide enough codes
16
for the mobile terminals to assure the ideal situation these are not used in practice. In application the network operators take into account the cost of equipment and the network load. They use Erlang B to assure certain grade of service. To overcome the shortcomings of the modeling presented herein, in future work the outage probability and the loading control in the planning process are need. An analysis considering outage probability and cell loading will give a more accurate relationship between the coverage and capacity. VI. NOMENCLATURE AND ELABORATION MSC AuC HLR PLMN RAN RNS N Mobile Switching Center Authentication Center Home Location Resister Public Land Mobile Network Radio Access Network Radio Network Subsystem Users in a cell Thermal Noise bit-to noise power density Data rate Voice gain Received power from mobile user EIR GGSN VLR RNC SGSN UTRAN S SIR Equipment Identity Register Gateway GPRS Support Node Visitor Location Resister Radio Network Controller Service GPRS Support Node UMTS Terrestrial RAN Signal Signal-to-interference ratio chip rate Gain factor Section gain The transmission power of the user
Eb / I o R Gv
S
W
G
Gs
S1
VII. REFERENCES [01] Harri H., Antti T., 2005, "WCDMA for UMTS Radio Access for Third Generation Mobile Communication". Wiley, pp. 120-143. [02] Jaana L., Achim W., Tomas N., 2002, "Radio Network Planning and Optimization for UMTS", Wiley, pp. 38-54 [03] Wacker A., Laiho-Steffens J., 1999, "Static Simulator for Studying WCDMA Radio Network Planning Issues", Proc.1EEEVICSpwgCarf. [04] Hoppe R., Buddendick G., 2001, "Dynamic Simulator for Studying WCDMA for Radio Network Performance", Proc. IEEE International Symposium on Personal, Indoor and Mobile Radio Communications, pp: 32-37. [05] Jukka L., Matti M., 2001, "Radio Interference System Planning for GSM/GPRS/UMTS", Kluwer Academic Publisher, pp.42-48. [06] Venugopal V., 1999, "The Coverage-Capacity Tradeoff in Cellular WCDMA Systems", IEEE Transactions on Vehicular Technology. Vol.48, No.5. [07] Dinan E., December 2006, UMTS Radio Interface System, Bechtel Telecommunications Technical Journal, Vol.1(1), pp:10-14. [08] Gilhousen K. S., 1991, "On the Capacity of a Cellular WCDMA System", IEEE Transactions on Vehicular Technology. Vol.40, No.2. [09] Leung K. K., Mastey W. A., 1994," Traffic Modelling for Wireless Communication Networks" IEEE Journal on Selected Area in Communication. Vol. 12, No. 8. [10] Kyoung K, 2000, "Handbook of WCDMA System Design, Engineering, and Optimization" Prentice Hall.
17
Md. Amiruzzaman1, Md. Mustafizur Rahman2 and Md. Mahbubul Alam Joarder3 ABSTRACT
PDA, Smart Phone or other resource constraint devices has limited wireless communication capability and memory to serve the video on demand service. In this work, we propose a protocol to support and Video on Demand service on resource constrained devices and share among them. Designed protocol was successfully implemented and experimented on PDAs. Experimental results show that proposed protocol could successfully serve the video on demand using low rate wireless communication and share the video with similar devices efficiently. Keywords: Multimedia Stream, Embedded Network Engine, RTCP Multicast, Video-on-Demand
I.
INTRODUCTION
Resource constraint devices such as Personal Device Assistants (PDA) or Smart Phones became popular day by day and opened a new research area for the electronics and telecom industries. These devices are accepted by the people as they can access to the digital world from anywhere they can be carried easily1. Despite of their constraints in resources, such devices are getting more and more advanced, including the ability to act as hard drives on computers via USB cables, and even merging with Smart-Phones. Now-a-days they offer wireless networking, remote storage, data compression, transferring files including the multimedia. However, audio and video streaming services could not be implemented due resource limitation and limited network capability. As a result, the PDA or Smart-Phone users cannot access and share video on demand (VOD) and online radio services2. Modern PDAs and Smart Phones support both IEEE 802.11 (WLAN) and IEEE 802.16e (WiMax and WiBro) protocols with limited functionality. Our work mainly focuses on Video on Demand sharing over Wireless Broadband (WiBro) transmission between PDAs, and also covers the client-server communications between PDA and Server. Proposed network engine considers the memory capacity of PDA, transmission rate, bandwidth, connection procedure and user interaction. It is implemented on HP iPAQ PDAs and experimented with some real video clip.
Figure 1. Proposed Video Sharing between PDAs Advances in computing and communication in recent years have made video on demand sharing feasible between resourceful devices like computer systems or digital media devices. In VOD systems, movie database is stored in a central
server or in a group of interconnected servers. The video server is connected with high-capacity network interfaces through which local distribution centers (hubs) receive the stream data and broadcast to the
households. Such systems can share video stream between them by high-speed networks and buffering in larger memories. However, similar VOD sharing between PDAs or Smart-Phone could not be offered due to PDA-to-PDA low rate wireless
18
Md. Saiful Islam et al.
transmission, insufficient memory for buffering, and absence of efficient communication tool. In the next section, we discuss some of the related works and in section III we present the proposed framework for VOD service in PDAs and sharing. Section IV describes the implementation issues of the VOD sharing followed by experimental results in Section V. Section VI concludes the paper.
II.
LITERATURE REVIEW
There is no remarkable work found on sharing video on demand for resource constraint device especially for PDA and Smart Phones. However, several works have been done on PC based video streaming engine and some of them can be considered in designing VOD service sharing in resource constraint devices. Jehan-Francois Paris et al. proposed a broadcasting technique for low bandwidth channels with multiple viewers3. Chiung-Shien et al. proposed a stream pumping engine for scalable Video-on Demand System and scalable video on demand server4-5. Lawrence A. Rowe et al. describes a strategic retreat to assess the current state of multimedia research and suggest directions for future research6. Some research works are also available on QoS issues7-8 for video on demand services in low bandwidth channels. III. PROPOSED MODEL In our proposed model, the clients register themselves for VOD access at the server using a client-server connection from the network engine. The video stream from the VOD server can also be shared with any of its neighboring PDA or the buddy list using a peer-to-peer connection. As the PDA to PDA wireless links is cannot support high data rate, and there is a very limited memory in both PDAs, we propose to establish a virtual connection between the PDAs During a video stream share, the source PDA transfers the same RTCP connection link to the destination PDA. Since the video servers delivers all packets as multicast packets and the wireless access point (WAP) broadcasts the packet within its transmission range, both sending and receiving PDAs receive the multicast packet and delivers to the application. Thus using this tricky methodology, we design the sharing protocol quite easy and there is no additional packet transmission in the wireless medium. Figure 1 below shows the video sharing technique between PDAs. The proposed model for VOD sharing comprises of two components: first, a client-server connection for user authentication and VOD database access from PDA to VOD server; and the second, is a peer-to-peer connection between two PDAs to share the Video stream with another PDA. IV. IMPLEMENTAION We implement the VOD network engine in HP iPAQ hw6945 Mobile Messenger with operating system Windows Mobile 5.0 Phone Edition. In contrast to the generic computing devices, programming tools in PDA environment offer limited functionalities. J2ME, embedded visual c++, embedded visual basic and embedded c (for Embedded Linux) are such programming compilers. Further, for rendering graphics OpenGL ES is used instead of OpenGL. In our work we have chosen the Embedded Visual c++ as our platform was Windows Mobile, for embedded Linux the Embedded C can be used.
Figure 2(a): Client-Server Connection Protocol
16
An Architecture of Active Learning SVMs with Relevance Feedback for Classifying E-mail
Figure 2(b): PDA-to-PDA Connection Protocol in case 1, when the source PDA is already receiving a video stream
Figure 2 (c) PDA-to-PDA Connection in case 2, when source PDA not receiving any video stream. PDA-VOD Server (Client-Server Connection): In implementing the PDA-VOD client server connection, we follow a three-phase connection: User Authentication, Video List transmission for selection and finally the video stream transmission. A Real-time Transport Control Protocol (RTCP) client socket is created up on request from the PDA application to initiate the User Authentication and billing at the VOD server. In the second phase, legitimate users request for the video list for selection and the server send the list to the PDA. Finally, when the PDA selects desired video program, the VOD server sends the multicast video frames one after one as stream. Figure 2(a) shows the PDA-Server connection functionality. We prefer to use multicast property of RTCP for a two-fold benefit: first to avoid the buffering at the PDA and second to avoid second transmission from the source PDA to the destination PDA during video sharing. PDA-to-PDA (Peer-to-Peer): We select a tricky method for the PDA-to-PDA connection for video sharing. As we propose to use the RTCP multicast protocol for video sharing, the repeated transmission from the source PDA to the destination PDA for each packet avoided, and thus, proposed method does not overwhelm the low rate wireless link. The PDA-to-PDA connection can be established in two ways for two cases. The first case is applicable when the source PDA has already been established a PDA-Server connection with the VOD server, i. e., when the source PDA is receiving a selected video stream from the server. The second type comes to the focus when the source PDA did not select any video. Two cases for the proposed video sharing protocol have been shown in figure 2(b) and 2(c). In case 1, when the source PDA already opened a client-server connection with the VOD server and it receives a sharing request from any of its buddies; the PDA can only share the on-going video transmission. The destination PDA (request PDA) opens a connection with the source PDA (sharing PDA) for video list. The source PDA then replies with single entry ongoing video information. If the requested PDA selects the entry, then the source PDA transfer the connection
17
Md. Saiful Islam et al.
link to the destination PDA and then it can receive the multicast packet. This case is shown in Figure 2(a). In case 2, when the sharing/source PDA has not created any client server connection with the VOD server, it opens a client server connection immediately after receiving the sharing request from destination PDA and receives the available video list and send the list to the destination PDA for selection. The destination PDA selects the video and via the source PDA it initiates the video stream transfer. Source PDA transfers the link to the destination PDA. V. RESULT ANALYSIS
The implemented protocol has been experimented with a number of test cases with variations in different transport layer packet sizes, and video compressions. Performance of the streaming service for both PDA-Server and PDA-TO-PDA connection has been observed. Table 1: PDA-Server Packet Transmission Delays Packet Delay Size No. of Loss Min. Max. Avg. (byte) Packets (%) (ms) (ms) (ms) 1024 50 0 7 209 66 2048 50 0 11 112 62 4096 50 0 19 117 69 8192 50 0 34 130 81 16384 50 0 59 157 107 32768 50 0 109 209 159 65535 50 1 212 329 264 Table 1 shows the observed per packet delay for video streaming in PDA-Server Connection and corresponding graph is shown in Figure 3. From the results, we find that the PDA-Server connection is capable of providing VOD service on PDAs. Only few packets are dropped with the maximum RTCP packet size (64KB). There is no packet loss below that packet size. Video streaming services generally use the 14KB, which is shown in shaded row in Table 1. Delay columns in Table 1 shows that 16KB packets can be delivered perfectly to cope up with the video information timing. In busy wireless network environment, larger packet sizes can experience larger delays that may degrade the video performance. Shaded region in Figure 3 shows the optimum tolerable video frame packet delay and we find that packet size of 14KB to 32KB the video frames appear in real-time.
Figure 3: Average RTCP packet delay In case of the PDA-TO-PDA connections for video sharing, 16 byte packets are suitable for data transmissions because no real-time multimedia is transferred using the connection. List of video and the connection link with the VOD server are transferred between them. Table 2 shows the observed delay summary for 16 bytes payload size in PDA-TO-PDA transfer. Table 2: PDA-to-PDA Packet Transmission Delays Delay Size No. of (byte) Packets Min. Max. Avg. (ms) (ms) (ms) 16 100 128 271 168
VI. CONCLUSION 18
We presented an embedded network-engine framework for Video on Demand service that is suitable for the PDA or other resource constraint devices with wireless communication environment. Proposed network-engine has been implemented on PDAs and the results we obtained from experiment justify its acceptance for resource constraint devices. The increasing popularity of VOD services to the users and the recent surge of PDAs and Smart Phones, we believe that such a system, which enables people to easily access video information anywhere and anytime, would be beneficial to many application areas such as education, entertainment, and business. Combining the developed network engine with real applications running in digital cameras and developing some data recovery techniques for large-scale VOD transmission are the prospective future work for this research. VI. REFERENCE [1] Tarik Taleb, Nei Kato, Yoshiaki Nemoto. On-demand media streaming to hybrid wired/wireless networks over quasi-geostationary satellite systems. Computer Networks: The International Journal of Computer and Telecommunications Networking. Volume 47, Issue 2. pp. 287 306. 2005. Video-on-demand service system and method of dynamic image distribution. European Patent EP1363456.http://www.freepatentsonline.com/EP1363456.html Jehan-Francois Paris, Steven W. Carter, and, Darrell D.E. Long. A Low Bandwidth Broadcasting Protocol for Video on Demand. Seventh International Conference on Computer Communications and Networks (ICCCN '98) p. 690, 1998. Chiung-Shien Wu Gin-Kou Ma and Bao-Shuh P. Lin On the design of stream pumping engine for scalable Video-on Demand System. IEEE serial number: 0-7803-3734-4/97, 1997, pp-72-73 Chiung-Shien Wu Gin-Kou Ma and Bao-Shuh P. Lin A scalable architecture for Video-on-Demand Servers, IEEE Transactions on Consumer Electronics, Vol. 42, No. 4, serial number: 0098-3063/97, Nov. 1996, pp1029-1036
[2] [3]
[4] [5]
Md. Saiful Islam1 and Md. Iftekharul Amin2 ABSTRACT
In this paper, we have proposed an architecture of active learning SVMs with relevance feedback (RF) for classifying e-mail. This architecture combines both active learning strategies where instead of using a randomly selected training set, the learner has access to a pool of unlabeled instances and can request the labels of some number of them and relevance feedback where if any mail misclassified then the next set of support vectors will be different from the present set otherwise the next set will not change. Our proposed architecture will ensure that a legitimate e-mail will not be dropped in the event of overflowing mailbox. The proposed architecture also exhibits dynamic updating characteristics making life as difficult for the spammer as possible. Keywords: Machine Learning, Artificial Intelligence, and Network Security, Support Vector Machines, Maximum Margin Hyperplane, Spam, Active Learning and Relevance Feedback.
19
I. INTRODUCTION Due to the rapid development of WWW and Internet, e-mail becomes a very efficient, convenient and shortcut way of communication. But at the same time despite e-mails numerous advantages, it has become the victim of abuse. The mass posting of unsolicited bulk commercial e-mail has become an increasingly large problem. This unsolicited and unwanted email is known as Spam in the literature. This increase in Spam is annoying. An interesting aspect of this rise in spam is that unique spam attacks are also on the rise. More about spam will be found in http://www.junke-mail.org, http://spam.abouse.net and http://cauce.org. There have been various learning machines to classify e-mail [1-3]. Solutions to the proliferation of spam includes either technical or regulatory [4]. Technical solutions include filtering based on sender address or header content. Microsoft junk and adult content e-mail filters work by looking for key words emphasizing on which words the filter will look for and where [5]. Although sender address based filtering may be useful but content based filtering may sometimes block valid messages [6-7]. Since missing legitimate e-mail is an order of magnitude worse than receiving spam, we have no intention to automatically reject e-mail that is classified as spam. Rather, we will allow user to mark (label) their e-mail as either spam or nonspam. After a finite number of examples are collected, the learning machine will trained and performance on new examples will be predicted. Our proposed mail server will deliver email to the user in decreasing order of probability that the e-mail is nonspam. It is then up to the user to either read the e-mail or trash the e-mail. He or she can also provide feedback to the mail server by indicating if an e-mail misclassified. This is what we want to define relevance feedback (RF). Then we will modify our classifier to take into account this. In the case of overflowing mailbox to provide rooms for less probable spam mail that has the high probability to be legitimate we will delete mails at the bottom of the list. This is one of the promising characteristics of our architecture. The paper is organized as follows: Section II. will describe recent trends and techniques used to classify e-mail using SVMs, Section III. will describe briefly about SVMs, Section IV. will describe active learning strategy, Section V. will describe several design choices and issues section VI. Will describe the proposed architecture and References are listed in Section VIII. II. RELATED WORKS The success of support vector machine (SVM) in solving real-life problems made it not only a tool for the theoretical analysis but also a tool for developing algorithms for real-world problems [1-2, 8-10]. Their efficiency in providing solutions to classification and function approximation problems is the ongoing research issue. Our intention is to classify e-mail using support vector machines. Since this is a new type of research in the area of machine learning, much of the works are still undone. Drucker et al. [2] have described SVMs for classifying e-mail and showed that SVMs perform best among different learning algorithms. They emphasized on rank ordering of the e-mail rather than rejecting it. But since their learner learns only once, it will not be able to cope with newer strategies adopted by spammer that can fake their classifier [7]. To take account into this problem our classifier will update itself upon receiving relevance feedback from users if any e-mail misclassified. Therefore it will be almost impossible for the spammer to misguide our classifier. The concept of relevance feedback we used here is as same as relevancy feedback (RF) in information retrieval (IR) of text documents [10]. The advantages of support vector machines with active learning strategy for spam classification has been described in [1]. But though their architecture exhibits active learning strategy, it is incomplete in the above sense that we claimed for [2]. In this paper we have described a complete architecture of active learning SVMs with relevance feedback for classifying e-mail that is free from of the above problems. It also presents other promising advantages that may prove it as a unique candidate. III. SUPPORT VECTOR MACHINES The key concepts of SVMs are the following: there are two classes, examples:
yi {1,1}
and there are N labeled training
( x1, y1), ( x 2, y 2),..., ( xN , yN ), x R d w

2
where d is the dimensionality of the vector. If the two classes are linearly separable, then one can find an optimal weight vector w such that is minimum; and
w xi b 1 if yi = 1; w xi b 1 if yi = 1;
or equivalently
yi ( w xi b) 1.
1.
Department of Mathematics, University of Dhaka, Dhaka-1000. Email: msa417@hotmail.com.
Md. Showkat Ali.
Training examples that satisfy the equality are termed support vectors. The support vectors define two hyperplanes, one that goes through the support vectors of one class and one goes through the support vectors of the other class. The distance between the two hyperplanes defines a margin and this margin is maximized when the norm of the weight vector
is
minimum. Vapnik has shown we may perform this minimization by maximizing the following function with respect to the variables j :
W ( ) = i ij ( xi xj ) yiyj
subject to the constraint: 0 j where it is assumed there are N training examples, xi is one of the training vectors, and represents the dot product. If j > 0 then xj is termed a support vector. For an unknown vector xj classification then corresponds to finding
i =1 i =1 j =1
F ( xj ) = sign{w xj b}
where
w = iyixi
i =1
and the sum is over the r nonzero support vectors (whose s are nonzero). The advantage of the linear representation is that w can be calculated after training and classification amounts to computing the dot product of this optimum weight vector with the input vector. For the nonseparable case, training errors are allowed and we now must minimize
w 2 + C i
i =1
subject to the constraint
yi ( w xi b) 1 0 W ( )
but the constraint is now
is a slack variable and allows training examples to exist in the region between the two hyperplanes that go through the is quadratic in and may be solved using quadratic programming techniques.
support points of the two classes. We can equivalently minimize
0 j . Maximizing W ( )
0 i C
instead of
The advantage of linear SVM is that execution speed is very fast and there are no parameters to tune except the constant C. Drucker et. al. [2] has shown that the performance of the SVM is independent of the choice of C as long as C is large enough (over 50). Another advantage of SVMs is that they are remarkably intolerant of the relative sizes of the number of training examples of the two classes. In most learning algorithms, if there are many more examples of one class than another, the algorithm will tend to correctly classify the class with the larger number of examples, thereby driving down the error rate. Since SVMs are not directly trying to minimize the error rate, but trying to separate the patterns in high dimensional space, the result is that SVMs are relatively insensitive to the relative numbers of each class. For instance, new examples that are far behind the hyperplanes do not change the support vectors. The possible disadvantages of SVMs are that the training time can be very large if there are large numbers of training examples and execution can be slow for nonlinear SVMs, neither of these cases being present here [2]. IV. ACTIVE LEARNING Pool based active learning involves selecting a training set of examples T from a pool of unlabeled examples U. The examples in T are labeled and used for learning. An iterative process involving active learning at each step is called the active learning cycle. This cycle proceeds as follows. At first, all of the examples in the pool are unlabeled and therefore members of U, the pool of unlabeled examples. An initial training set To must have at least one active and one inactive example in order for the maximum margin hyperplane to be found. Therefore we generate To by randomly selecting examples from U and labeling them until at least one example with each label is in To. We then remove the new examples from the unlabeled pool by making Uo = U To. In turn, we can use

20
A Note On Symplectic And Contact Geometry
active learning to find a new training set T1. T1 contains all the examples from T0, using some selection criterion, we can choose a batch of N examples from Uo, label them and place them in set T1. And we get U1 = U T1. Repeat this process we can find training set T2 and so on. The iteration of active learning process is: Step1. To find To by randomly selecting examples in U, until one of each label is found. Uo = U To. Step 2. In iteration i, select a batch Bi of N examples from Ui-1 using some selection criterion. Step 3. Find the label of each example in Bi. Step 4. Ti = Ti-1 Bi, Ui = U Ti. Step 5. Return to step 2 and repeat from step 2 to step 4. Here we are first given a set of both labeled and unlabeled data. The pool of training examples can be colleted from individual inboxes, thus it will almost impossible for spammer to write spam messages that can beat our e-mail filter [7]. V. DESIGN CHOICES 1. Feature Representation A feature is a word. In the development below, w refers to a word, x is a feature vector that is composed of the various words from a dictionary formed by analyzing the documents. We take a word as a feature only if it occurs in three or more documents. There is one feature vector per message. w refers to a weight vector usually obtained from some combination of the xs. There are various alternatives and enhancements in constructing the vectors [2]. We will use the TFIDF. TF-IDF uses the TF (term frequency) multiplied by the IDF (inverse document frequency). The document frequency (DF) is the number of times that word occurs in all the documents (excluding words that occur in less than three documents). The inverse document frequency (IDF) is defined as
D IDF ( wi ) = log DF ( wi )
where
D is the number of documents. Typically, the feature vector that consists of the TF-IDF entries is normalized to unit
length [2]. This kind of representation scheme leads to a very high dimensional feature spaces containing 10000 dimensions or even more [2]. SVMs use overfitting protection mechanism, which does not depend on the number of features. They have potential to handle this kind of problems [1]. Use of stop list consisting of most frequent words like of, and, the, etc. and punctuation marks ., ,, ; etc. can be used to lessen the dimensionality [2]. But it is observed that there are more punctuation marks in the legitimate e-mail than in the spam [7]. This indicates that spammers are less interested in good punctuation than the other people. Looking at the most frequent words, it is not obvious which e-mail is spam and which is legitimate. Therefore we concentrated on using all the features rather than a subset. 2. Performance Criteria Error Rate: Error rate is the typical performance measure for two-class classification schemes. However, two learning algorithms can have the same error rate, but the one which groups the errors near the decision border is the better one. False Alarm and Miss Rate: We define the false alarm and miss rates as
The advantage of the false alarm and miss rates is that they are a good indicator of whether the errors are close to the decision border or not. Given two classifiers with the same error rate, the one with lower false alarm and miss rates is the better one. 3. Training Examples The pool of training examples consists of the individual inboxes. Initially there will be no classifier running. Users will receive e-mails whether it is spam or nonspam. But after a certain amount of time users will be requested to send both legitimate e-mail and spam they considered. This colleting process will be continued until pool is sufficient for the learner. When the learning procedure will complete from training examples, our classifier will be activated and users will receive e JournalofComputerScience,Volume1,Number1,June2007,ISSN:19946244
21
Md. Showkat Ali.
mails according to the probability of their e-mail of being spam. Selection Methods: Here we are given a set of both labeled and unlabeled data. The learning task is to assign labels to the unlabeled data as accurately as possible. To construct SVMs, we must construct an optimal classification hyperplane, i.e. the maximal margin hyperplane. Assume the training data: one subset I for which
( x1, y1), ( x 2, y 2),..., ( xN , yN ), x R d , yi {1,1} y =1, and another II subset for which y =-1can be separated by a hyperplane.
If the pattern datasets can be separated without error, the Euclidian distance between the support vectors and the hyperplane must be the largest, i.e. the margin. The maximal margin classifier for our task can be build based on the following theorem.
U = {( x1, y1),..., ( xN , yN )} the hyper plane that solves the problem min ( w w) subject to yi (( w xi ) b) 1, i = 1,...N
realizes the maximal margin hyperplane with geometric margin
Theorem: Given a linearly separable training examples
= 1
2.
There are few criteria that can be used in each iteration of the active learning cycle as mentioned above. One method is to choose the closest examples to the maximum hyperplane. In step j we select the closest examples to the maximal margin hyperplane of T j . But it will be difficult to perform active learning if there are no unlabeled examples closer to the hyperplane than the support vectors. Another selection method is the furthest selection algorithm, which chooses the furthest examples from the hyperplane. This method is not theoretically motivated, and we expected to perform it poorly.
Fig 1. A maximum margin hyperplane with its support vectors VI. Overview of Architecture The block diagram of our proposed architecture is shown in figure-2. The architecture includes a user interface that is responsible for communicating with the user, a pool of training examples, active learning module, e-mail classifier and user mailbox. The total module can be in either Training mode (TM) or Active Mode (AM). In training mode the module will collect individual user e-mails that they considered as spam or legitimate. After collecting sufficient amount of both spam and legitimate e-mails the learning process will begin. Now the module will switch to active mode. In this mode the module classify users e-mails as spam or nonspam and place them in his or her mailbox according to their probability. When a user observes that an e-mail is misclassified, he or she will provide feedback with misclassified e-mail and its label he or she considered. After receiving feedback the module will switch to training mode and above described procedure will begin again.
22
A Note On Symplectic And Contact Geometry
Fig 2. Architecture of Active Learning SVMs with Relevance Feedback VII. SUMMARY AND FURTHER WORKS The paper described an architecture of active learning SVMs with relevance feedback for classifying e-mail. We have theoretically given reason why our provided architecture will be unbeaten from spammers. We are now bending ourselves in simulating our architecture and its practical implementation. Although a direct comparison has not been made here, the results of SVMs for spam categorization [2] made it promising. The paper actually motivated to find a framework for spam filter using SVMs. It is a try to explore dynamic updation in learning module that seems not to be written again due to change in spam. VIII. REFERENCES
[01] L. Kunlun and H. Houkuan An Architecture of Active Learning SVMs for Spam, ICSP Proceedings, 1247-1250, 2002. [02] H. Drucker, D. Wu, V. N. Vapnik, Support Vector Machines for Spam Categorization, IEEE Transactions on Neural Networks, Vol. 10, No. 5, Sept. 1999. [03] W.W. Cohen, Learning rules that classify e-mail, In Proc. AAAI Spring Sump. Inform. Access, 1996. [04] L. Faith, Craner and B.H. LaMacchia, Spam!, Commun. ACM, Vol. 41, No. 8. pp 74-83, August 1998. [05] http://office.microsoft.com/en- us/assistance/HP010421791033.aspx [06] Stefanie Olsen, Staff writer, CNET News.com, AT&T spam filter loses valid e-mail, Published: January 24, 2003, 5:21PM PST. [07] Cormac OBrien, Compiling a Corpus of E-mail to Evaluate Spam Filters, https://www.cs.tcd.ie/Cormac.OBrien/ecai2002.pdf [08] X. Wang and Y. Zhong, Statistical Learning Theory and State of the Art in SVM, Proceedings of the second IEEE international Conference on Cognitive Informatics (ICCI), 2003. [09] M.P.S. Brown, W.N. Grundy, D. Lin, N. Cristianini, C. W. Sugnet, T. S. Furey, M. Ares, Jr. and D. Haussler, Knowledge-based analysis of microarray gene expression data by using support vector machines, PNAS, Vol. 97, No. 1. 262-267, Jan. 4, 2000. [10] H. Drucker, B. Shahraray and D. C. Gibbon, Relevance Feedback using Support Vector Machines, In Proceedings of the 18th International Conference on Machine Learning, pages 122--129, 2001. [11]
[6]
Lawrence A. Rowe, Ramesh Jain. ACM SIGMM retreat report on future directions in multimedia research. ACM Transactions on Multimedia Computing, Communications, and Applications (TOMCCAP) archive, Volume 1 , Issue 1. pp. 3 13, 2005 Baschieri, F. Bellavista, P. Corradi, A. Mobile agents for QoS tailoring, control and adaptation over the Internet: the ubiQoS video on demand service. In Proceedings of the Symposium on Applications and the Internet, ISBN: 0-7695-1447-2, pp. 109-118. 2002. Michael Ditze, Chris Loeser, Peter Altenbernd. Improving content replication and QoS in distributed peer-topeer VoD appliances. In Proceedings of the 24th International Conference on Distributed Computing Systems Workshops. pp.78- 84. 2004.
[7]
[8]
23
Md. Showkat Ali.
A Note on Symplectic and Contact Geometry

Md. Showkat Ali1
ABSTRACT
Contact geometry is the odd-dimensional twin of simplistic geometry. The connection between them is similar to the relation of projective and affine geometry. In this paper some new approach to simplistic and contact geometry are studied. Finally a main theorem in complex contact manifold in contact geometry is established. Keywords: Symplectic Manifolds, Contact Manifolds.
I. INTRODUCTION Contact geometry was born more than two centuries ago in the work of Huygens, Hamiltona and Jacobi as a geometric language for optics. It was soon realized that it has applications in many other areas, including non-holonomic mechanics and thermodynamics. One encounters contact geometry in everyday life when parking a car, skating, using a refrigerator or watching the beautiful play of light in a glass of water. Sophess Lie, Elie Cartan, Darboux [1] and many other great mathematicians devoted a lot of their work to this subject. However, till very recently most of the results were of a local nature. With the birth of symplectic and contact topology in the eighties, the subject was reborn, and the last decade witnessed a number of break through discoveries. There they found new important interactions with Hamiltonian mechanics, simplistic and sub Riemannian geometry, foliation theory, complex geometry and analysis, topological hydrodynamics, 3dimensional topology, and knot theory. Simplistic geometry is the mathematical apparatus of such areas of physics [2] as classical mechanics, geometrical optics and thermodynamics. Simplistic geometry is the geometry of a closed non-degenerate 2-form [3] on an even-dimensional manifold [4]. Contact geometry is the geometry of a maximally non-degenerate field of tangent hyper planes on an odd-dimensional manifold. The symplectic structure is fundamental for Hamiltonian dynamics and in this sense symplectic geometry is as old as classical mechanics. Symplectic manifold are an intermediate case between && real and complex (K a hler) manifolds. Symplectic manifolds still play an important role in recent topics in physics, such as string theory. The contact manifolds are the odd-dimensional analogues of symplectic manifolds. In this paper a theorem on complex contact manifolds in contact geometry is focused that generalizes on symplectic manifolds in symplectic geometry.
II. PRELIMINARIES In this part, we will give some definitions in vector spaces and prove some Propositions. 1. A symplectic vector space is a pair (V , ) consisting of a finite-dimensional real vector space V and a non-degenerate skew-symmetric bilinear form (i)
: V V . This means that the following conditions are satisfied:
Skew-symmetry: for all
u , v V , (u , v ) = (v, u ).
(ii) Non-degeneracy: for every u, v V , (u , v) = 0, v V u = 0. The vector space V is necessarily of even dimension since a real skew-symmetric matrix of odd-dimension must have a kernel. A linear symplectomorphism of the symplectic vector space (V , ) is a vector space isomorphism : V V which preserves the symplectic structure in the sense that
* =

24
Measurement of Specific Electrical Resistance of Nine Pure Liquid Metals at High Temperature ( 15000C) using Rotating Magnet-field Apparatus.
where
* (u , v) = ( u , v)
for u , v V . The linear symplectomorphisms of (V , )
form a group [5] which we denote Sp (V , ). In the case of the standard symplectic structure on Euclidians space we use the notation Sp ( 2n) = Sp ( R2n, where
0 )
j =1 2. Let (V , ) be a symplectic vector space and F V be a subspace. The symplectic complement of F is the subspace F = { X V | ( X , Y ) = 0, Y F } .
The properties of the symplectic complement are summarized in the following: If F and G be the subspaces of a symplectic vector space (V , ) , then
0 =
dx dy . j j
F whenever F G , (ii) ( F ) = F ,
(i) G
(iii)
( F + G) = F G , (iv) ( F G ) = F + G , (v) dim F = dim V dim F .

3. A subspace F of a symplectic vector space V is said to be
F F, (ii) coisotropic whenever F F , (iii) symplectic whenever F F = {0} , (iv) Lagrangian whenever F = F .
(i) isotropic whenever
F is coistropic and conversely. It follows from the above properties that a Lagrangian 1 subspace of V is the same as an isotropic subspace of dimension dim V . The categories are not exhaustive, a general 2
Note that if F is isotropic then subspace is neither isotropic, nor coisotropic, nor symplectic. Proposition
1 2
1.
n
Let
(V , )
be
2n-dimensional
symplectic
vector
space.
Then
has
basis
{ X , X , L , X , Y1 , Y2 , L , Yn } such that
a ( X a , X b ) = 0, 2 ( X a , Yb ) = b ,
(Ya , Yb ) = 0
where a, b = 1,2, L , n and such a basic { X , Yb } is called symplectic frame.

a
To proof this Proposition, let Q be a Lagarngian subspace of V and let W be some other n-dimensional subspace such that
V = W Q . Then W identifies W with Q * by mapping Z W to the linear form 2 ( Z ,) Q * . 1 2 n * Let {Y1 , Y2 , L , Yn } be a basis for Q and let {Z , Z , L , Z } be the dual basis for W = Q . Then (Ya , Yb ) = 0 and
2 ( Z a , Yb ) = a . Put ab = ( Z a , Z b ) and X a = Z a + ab Yb (with the summation convention).

b
Since
ab = ba , we have
25
Matiur Rahaman Mian
( X a , X b ) = (Z
+ ac Y c , Z
+ bd Y d )
= ( Z a , Z b ) + ( Z a , bd Y d ) + ( ac Y c , Z b ) + ( ac Y c , bd Y d ) = ab + bd ( Z a , Y d )
ac ( Z b , Y c ) + 0
= ab + = ab = 0 1 ba 1 ab 2 2 1 ab 1 ab 2 2
and
2 ( X a , Yb ) = ( Z a + ac Yc , Yb ) = ( Z a , Yb ) + (ac Yc , Yb ) = ba + ac (Yc , Yb ) = ba + 0 =0
Therefore, { X , Yb } is a symplectic frame. Hence completes the proof. Proposition 2. Let F be a subspace of a symplectic vector space (V , ) and let onto a bilinear form
a
V = V /( F F ). Then projects
on
V and (V , ) is a symplectic vector space.
For the proof of this Proposition, let
: F V
denote the projection along
F F . Define by
( X , Y ) = ( X , Y ) where X , Y F and X = ( X ) and Y = (Y ) . This is clearly well-defined, skew-symmetric and non-degenerate. Hence (V , ) is a symplectic vector space. Hence completes the proof.
III. SYMPLECTIC AND CONTACT MANIFOLDS 4. A symplectic manifold is a pair ( M , ) consisting of a manifold M of dimension 2n and a differential 2-form such that (i) is closed, i.e., d = 0, (ii) is non-degenerate, i.e., x M and any v x Tx M , one has (v x , wx ) = 0, wx Tx M if and only if
v x = 0. x1 , x 2 , ... , x n , y1 , y 2 , ... , y n equipped with a differential
If we consider the space M = R2n with coordinates form =
dx
i =1
dy i . Then the pair ( M , ) is a symplectic manifold.
Note that the above manifold M is called complex symplectic if M is a complex manifold and
is a holomorphic 2-form.
1. Darboux Theorem 1. (For symplectic Manifold) Let ( M , ) be a 2n-dimensional symplectic manifold and p be any point in M. Then there is a coordinate chart
(U , x1 , x 2 , ... , x n , y 1 , y 2 , ... , y n ) centered at p such that on U,
= dx i dy i
i =1
26
The chart
(U , x1 , x 2 , ... , x n , y 1 , y 2 , ... , y n ) is called a Darboux chart.
5. A contact manifold is a pair (Y , D ) consisting of (2n + 1)-dimensional manifold Y and a rank 2n-subbundle D TY of the tangent bundle to Y which is maximally non-integrable in the sense that the Frobenius form [6]
: DD TY /D
is everywhere non-degenerate. The condition of non-integrability has a very useful reformulation, which we shall now describe. Given a co-dimension 1 subbundle D TY , let L := TY / D denote the quotient line bundle on Y, we then have an exact sequence 0 D 2 n TY 2 n +1 L 0 , where is the tautological projection and D = ker . But we may also think of as a line bundle-valued 1-form H 0 (Y , 1Y L) , and so attempt to form its exterior derivative d . Unfortunately, this ostensibly depends on a
(u , v ) a [u , v] mod D
choice of local trivialization; for if
is any 1-form, d ( f ) = f d + df . However, it is now clear that

*
d D is
well-defined as a section of L D , and an elementary computation shows that the Frobenius form mentioned above. Now if the skew form 2n, so that Y must have odd dimension
d D thought of in this way, is exactly
2n + 1 3 . Moreover, the non-degeneracy condition exactly requires that [6]
d D is to be non-degenerate, D must have positive even rank
( d ) n 0 .
Proposition 3. The map : D D TY / D is a OY linear (or a tensor). For the proof of this Proposition, by the definition of Frobenius form Now consider for all local function f OY, we have
( u , v ) = [ u , v ] mod D L.
( fu , v) = [ fu , v] mod D
= = = f [u , v] mod D v( f )u mod D (Since u D and v ( f ) is a function, therefore, v ( f )u D so that f [u , v] mod D 0 f (u , v).
v ( f )u mod D vanish).
( u , v ) = ( v ,u ) . So we can conclude that ( u , fv ) = f ( u , v ). Therefore the map is a OY linear or a tensor. Hence completes the proof.
Note that If we consider
M n be a manifold and T * M its cotangent bundle. Then the projectivization of T * M , P (T * M ) is a contact manifold whose dimension is 2 dim M 1 i.e., 2n-1.
2. Derboux Theorem 2. (For contact Manifold) [7] Every differential 1-form defining a non-degenerate field of hyperplanes on a manifold of a dimension (2n + 1) (contact manifold) can be written in some local coordinate system in the normal form
= xdy + dz
where
x = ( x1 , x 2 , ... , x n ), y = ( y1 , y 2 , ... , y n ) and z- the local coordinates.

IV. THEOREM ON COMPLEX CONTACT MANIFOLDS
27
Matiur Rahaman Mian
* 6. Let M be a complex manifold of complex dimension 2n + 1. Let contact manifold if the following conditions are satisfied: (i) On each
{U i } be an open covering of M. We call M a complex
U i there exists a holomorphic 1-form i such that i ( d i ) n is different from zero at every point of
Ui .
(ii) If
U i U j is nonempty, then there exists a nonvanishing holomorphic function f ij on U i U j such that i = f ij j on U i U j .
1. Theorem 3. (Main theorem) If M is a complex contact manifold of complex dimension 2n + 1, then (i) The structure group of the tangent bundle of M can be reduced to
U (1) ( Sp ( n) U (1)).
(ii) Let ci (M ) be the i-th Chern class of (the tangent bundle of) M and characteristic class of the line bundle over M defined by
the
{ f ij } . Then
1 + c1 ( M ) + c 2 ( M ) + L L =
(1 + )(1 + n + L L)
In particular, c1 ( M ) is divisible by n + 1 ; c1 ( M )
= (n + 1) .
that P is a real contact
(iii) There exists a principal fiber bundle P over M with structure group U (1) such
manifold. Moreover, both P and a real contact form on P can be constructed in a natural way from M and
i .
It has been shown in [2] that the structure group of the tangent bundle of an orientable real contact manifold of real dimension 2n+1 can be reduced to SO (1) U ( n)( = SO (1) ( SU ( n) U (1)). To proof of (i), let
Tx (M ) be the complex tangent space to M at a point x. Assume x to be in U i . As i 0 at x, i = 0 defines a 2n dimensional complex vector subspace Fx of Tx (M ) . Let F be the vector bundle over M with fibers Fx . Let E be the line bundle T ( M ) / F . Then, T ( M ) F E . From the definition of contact form, it follows that d i , on Fx ,
is of maximal rank and is defined up to a factor, i.e.,
d j Fx = f ji d i Fx .
Now (i) follows immediately from the definition of Sp ( n) . For proof of (ii), let us consider
i = f ij j , it follows easily that
i (di ) n = ( f ij ) n +1 j (d j ) n .
Since
i (d i ) n is a holomorphic form of degree 2n + 1, { f ij ( n +1) } defines the canonical bundle K of M. The

{ f ij } . From K = E ( n +1) , we
characteristic class of K is c1 ( M ) . The line bundle E = T ( M ) / F is defined by obtain
c1 ( M ) = (n + 1) .
1 + c1 ( M ) + c 2 ( M ) + L =
Let ci (F ) be the i-th Chern class of the vector bundle F. By the Whitney Duality Theorem,
From c1 ( M )
= (n + 1) , it follows that c1 ( F ) = n . Hence completes the proof.

(1 + )(1 + c1 ( F ) + c 2 ( F ) + L)
28
Finally for the proof of (iii), let L be the line bundle over M defined by
{ f ij 1} and p the projection of L onto M. Let
hi : p 1 (U i ) U i C be the coordinate map. If v L x and x U i U j , then

hi (v) = ( x, z i ),
Hence
*
h j (v) = ( x, z j ),
z i = f ji z j .
z i . p * (i ) = z j . p * ( j ),
showing that { z i . p ( i )} defines a holomorphic 1-form
on L. A simple calculation shows that
( + ) (d + d ) 2 n +1 = A( z i z i ) n ( z i dz i z i dz i ) p ( i (d i ) i (d i ) )
We define a bundle P as follows. Let
* n n
where
A = (2n + 1)! /(n!) 2 .
B x be a positive definite Hermitian form on Lx differentiable with respect to x. Let Px = {v L x ; B x (v, v) = 1}, i.e., Px is a circle in L x . Then P = U xM Px is a principal fibre bundle over M with group U (1). We shall show that the restriction of + to P defines a real contact structure on P. If Px is given by
z i z i = b( x) 2 ; (b( x) > 0).
As
i (d i ) n i (d i ) n db is a form of degree 4n + 3 (> real dimension of M) on M, it vanishes identically.

( + ) (d + d ) 2 n +1 to P is
Hence, the restriction of
2 Ab 2n z i dz i p * ( i (d i ) n i (d i ) n ), which is clearly different from zero at every point P.
V. DISCUSSIONS In this paper we will provide two examples of proof of (ii) and (iii) for the Main Theorem. * 1. Complex projective spaces of odd dimension: Let complex vector space C
2n + 2
z 1 , z 2 , ... , z 2 n +1 , z 2 n + 2 be a coordinate in the (2n + 2)- dimensional

2n + 2
and let P 2 n +1 (C ) be the (2n + 1)- dimensional complex projective space. Then, C
-{0}
is the principal fibre bundle associated with a line bundle L over P 2 n +1 (C ) . Set
= z1dz 2 z 2 dz1 + L + z 2n+1dz 2n+ 2

z 2n+ 2 dz 2n+1
Let
{U i } be an open covering of P 2 n +1 (C ) and si a holomorphic cross section of the principal bundle C 2n + 2 -{0} over
*
4n+4
U i. Set i = si ( ). Then { i } defines a complex contact structure on P 2 n +1 (C ) . Considering C 2n + 2 as the real (4n
+ 4)-dimensional vector space R
, we obtain the (4n + 3)-dimensional real projective space P 4 n +3 (P), P 4 n +3 (R) is a principal fiber bundle over P 2 n +1 (C ) with a structure group U (1). Every odd dimensional real projective space is a real contact manifold. The standard real contact form on P 4 n +3 (R) is the one derived from the contact form of P 2 n +1 (C ) in the
manner described in the proof of (iii). * 2. Complex projective co-tangent bundles: Let V be a complex manifold of dimension n +1 and
~ on the dual complex tangent bundle T (V ) (= the space of complex co-tangent vectors) defined by ~ u Tv (T (V )) (u ) = v ( (u )), ~ ~ where is the projection of T (V ) onto V and is the differential of ; : T (T (V )) T (V ) . In terms of local ~ 0 1 n coordinate z , z , ... , z of V and the induced coordinate z 0 , z1, ... , z n , 0 , 1, ... , n of T (V ) ,
a holomorphic 1-form
29
Matiur Rahaman Mian
= 0 dz 0 + 1 dz 1 + 2 dz 2 + L + n dz n .
Fibre is the n-dimensional complex projective space (complex projective co-tangent bundle). This bundle of complex dimension 2n + 1 is our M. Considering the fibre of T (V ) as the (2n + 2)-dimensional real vector space, we take as P the cotangent sphere bundle over V (i.e., the fiber of P is a sphere in the fibre of T (V )). T (V ) V is the principle fibre bundle associated with a line bundle L over M. The definition of the complex contact structure on M is similar to the one in the first example. The classical real contact structure on the co-tangent sphere bundle P is the one derived from the complex contact structure on the complex projective co-tangent bundle M as described in the proof of (ii). VI. REFERENCES [01] [02] [03] [04] [05] [06] [07] Arnold, V. I. and Novikov, S. P., 1990, Dynamical Systems IV, Springer-Verlag, Berlin Heidelberg. Franel, T., 1997, The Geometry of Physics, Cambridge University Press. Darling, R. W. R., 1994, Differential Forms and Connections, Cambridge University Press. McDuff, D. and Salamon, D. A., 1998, Introduction to Symplectic Topology, Oxford University Press, New York. Libermann, P. and Marle, C-M., 1987, Symplectic Geometry and Analytical Mechanics, D. Reidel Publishing Company. LeBrun, C., 139(1), 1991, 1-43, Thickenings and conformal gravity, Comm. Math. Phys. Arnol'd, V. I., 1978, Mathematical Methods of Classical Mechanics, Springer Verlag.
Measurement of Specific Electrical Resistance of Nine Pure Liquid Metals at High Temperature (~15000C) using Rotating Magnet-field Apparatus.
Matiur Rahaman Mian1,PhD
ABSTRACT
Specific electrical resistance of nine pure molten metals, mercury, tin, bismuth, cadmium, lead, zinc, antimony, aluminum and silver were measured at different high temperatures using rotating magnet-field apparatus. The results for pure metals obtained were much the same with those of previous investigations.
Keywords : electrical resistance, high temperature, rotating magnet-field and apparatus. I. INTRODUCTION Many investigations have been already made of the electrical resistance of molten metals, especially in connection with the discontinuous change of electrical resistance during the molten of the metals. L. de la Rive [01] first studied the electrical resistance of six metals, namely tin, bismuth, cadmium, lead, zinc and antimony at high temperatures and found that in case of tin, zinc, cadmium and lead, the value of the electrical resistance increases abruptly during fusion, but in the case of antimony and bismuth, on the contrary, decreases abruptly during melting. G Vincentini and D Omodei [02] measures the electrical resistance of five metals namely tin, bismuth, thallium, cadmium and lead in the molten state and found that the specific resistance of molten metals at the melting point is proportional to the atomic weight. E F Northrup [03] studied the resistance of twelve pure metals, namely sodium, aluminum, potassium, copper, zinc, cadmium, tin, antimony, gold, quicksilver, lead and bismuth, at high temperature, and found that the electrical resistance of molten metals increases linearly with the rise of temperature in a wide range. H. Tsutsumi et al [04], Vassura et
30
A Comparative Study of Cell Based and Tree Based Call Admission Control for Mobile Cellular Network
al [05], Mller et al [06], Hackspill et al [07], Bridgman et al [08] and Bernini et al [09] also studied the resistance of molten metals and their results are much the same. II. EXPERIMENTAL TECHNIQUES The electrical resistively of the liquid alloys has been measured by several experimental arrangements which are in principle are based on the following two different methods [10]: 01. Direct method 02. Indirect (Electroden loss) Method: (a) Inductive Coupling Method (b) Rotating Magnet field Method The direct method is similar to the current-potential methods used with solid specimen. This method involves the introduction of thin capillaries [11], in which the potential drop along a capillary filled with the liquid is measured and converted into electrical resistivity by means of an accurate knowledge of the dimensions of the capillary. This method is in principle, capable of measuring very high precision, but in practice the method is suitable for above 2000C temperature. This is due to the experimental technique and thermoelectric e.m.f. Also in this method, most metals used to form connections to the melt because of their resistance to attack by liquid metals from thin rectifying oxide films, even at low temperatures and oxygen pressure that are impossible to reduce in hydrogen atmosphere. The presence of the films make precise measurement of the potential drop along the capillary very difficult and sometimes cause large discrepancies between values found with the different magnitudes or directions of the capillary current A second method involves an H cell with two large vessels connected by a thin capillary. The current and potential leads are contained in reservoirs. Matuyama [12] used a single capillary of length 300 mm and diameter 5.5 mm tube containing iron lid inserted in the surrounding large vessel of liquid. The absolute measurement of resistance in this method is not very accurate. The inductive coupling method developed by Nyberg and Burges [13] and Tomlinson and Licter [14] was the first scientist to use that method for the determination of electrical resistivity of liquid metals. The technique involves placing the sample in a solenoid producing an alternating magnet field, which induces eddy current in the sample. The observable effects are the change of impedance of the solenoid, more specifically, an increase in resistance and decrease in inductance of the solenoid, which is related to the sample conductivity (resistivity). In the four-prove method, the voltage drop along a well-defined specimen is measured using d.c current and the resistivity is calculated by Ohms law [15]. For higher temperatures, the rotating magnet-field method is more reliable. Several variants of this electrodeless method have been developed which are not always generally applicable. In the high-frequency method, the sample is dropped through a field of 10 MHz, the damping of the field being a measure of the electrical resistivity [16]. The rotating magnet field method was developed by Braunbeck, et al [17] and later modified by [18-23]. In this method, the specimen, in the form of a short cylinder, is suspended with its axis vertical in a magnet field rotating about the axis of the specimen. The eddy currents induced in the specimen interact with the applied field to produce a rotation of the specimen about the vertical axis. This rotation is measured and converted into electrical resistivity. This method is suitable for both solid and liquid specimens III. APPARATUS FOR MEASUREMENT OF RESISTIVITY The electrical resistivity was measured by the electroden loss Rotating Magnet-field Techniques developed by [17] used in this investigation. In the Soviet Union, the method was independently developed by Reget and described in details by Glazev et al [24]. The general principles of this techniques is further modified by Grube and Spiedel [18], Roll et al [19], Regel [20], Loffe and Regel [21], Samarin [22], Takeuchi and Endo [23].
31
Md. Shahid Hossain et al.
IV. THE APPARATUS The apparatus is illustrated in figure-1 The main furnace is a vacuum tight aluminum tube of length 333 mm, internal diameter 20 mm and outer diameter 27 mm is directly heated by a bifilar wound with Mo-resistance wire ( d = 5 mm) and opens at both ends. The central 250 mm is surrounded by an austenite (25-20 Cr-Ni) stainless steel tube, which carries the furnace winding, insulated from the tube by a layer of mica. The two tubes are separated by a gap of 1.06 mm and are held concentric by a pyrophyllite plug at each end of the outer tube. The furnace winding of 0.0359 inch in diameter tray in three non-inductive separate sections, each having a resistance of about 10 and each externally shunted by a variable resistance. The stainless-steel tube is surrounded by alumina brick and powder insulation contained inside 220 mm diameter asbestos of outer tube. The heating of the magnetic coil is prevented by radiation shields, by a vacuum of 10-2 Pa, and by water cooled outer shell. The rotating field is provided by a stator coil from a three phase stabilized alternating current regulator. The current intensities of the three phases can easily measured by a digital multi-meter. The field in the centre is provided approximately by H = 10i, where i is the coil current (max.30A) and H is field strength in gauss. The stator coil surrounds the asbestos outer tube of the furnace, which is approximately 225 mm in diameter. The 18 mm gap between stator and tube contains copper cooling coils which serve to maintain the stator at constant working temperature. The very weight of the stator is supported on three Duralium legs which are bolted to the brass base-plate. The base-plate is rigidly tied to the top-plate by the three tension brass rods. When assembled within its Dexion frame, the apparatus may be leveled by the three screws. The stator current is controlled by a three-phase voltage stabilizer and a variac auto-transformer. The currents are measured to 0.1% accuracy. The currents in all the three phases are adjusted to exactly equal. As the temperature coefficient of resistivity of liquid metals is usually about 5 x 10-8deg C-1. The control of furnace temperature is very important. A standard saturable-reactor type controller was used with a platinum resistance thermometer. The resistance thermometer is placed against the central furnace winding and can control furnace temperatures up to 15000C. The temperatures are measured with a calibrated Pt-Pt/13%Rh thermocouple contained in an alumina sheath whose tip lies about 2 mm below the suspended crucible. Thus a zone of 20 mm length with a temperature variation of 2oK could be obtained. The dimensions of the specimen are measured over the complete range of temperatures to be used for the resistivity measurements. Two probes, one of which can be moved in a vertical position, are placed in the liquid metal contained in a crucible in an inert atmosphere. They are connected electrically to a battery so that a current flows between them when both are immersed in the liquid. The liquid surface may then be located by withdrawing the sharp-pointed movable probe until the current ceases to flow. The probe is then again lowered until contact is made, and the true position of the surface recorded as the mean of the two positions of the probe, measured by the cathetometer. The bottom of the crucible may then be located exactly as the lowest point attainable by the probe. The difference between the two readings, corrected for the expansion of the probe, is then the correct length of the specimen
32
The specimen was placed in a pure Boron Nitride (BN) crucible. The boron nitride is inert, has a low electrical conductivity, low thermal expansion coefficient ( lin = 0.9 x 10-6 ) and can be easily machined. The BN crucible consists of internal diameter 14.3 mm and height 20.7 mm is shown in figure-2. The crucible is screwed and fitted with an additional BN suspended rod reaching into the cooler part of the furnace which carried at the upper end of the hexagonal concave mirror and the aluminum vanes. The mirror reflects a He-Ne Laser light beam through the optical window on to a linear scale of rigidly fixed distances of one meter. The caps also contain a small hole to allow the escape of gas during evacuation. The suspension Phosphor-bronze wire of length 200 mm and outer diameter 0.15 mm is attached at its upper end to a brass rod which enters the chamber through an O-ring seal, and which can rotate to zero the image on the scale or moved vertically to adjust the height of the crucible. The tungsten wire is clamped at either end by means of two chucks. The rotational movements of the suspension are damped by two arms which dip into a bath of silicone oil. Thus the angular movements of 0.01 radian can easily detect. The entire torsion system (Crucible, BN-rod, Hexagonal mirror, Al-vanes) is suspended by the wire. The torsion angle can be determined with a He - Ne Laser light beam by measuring the distance on two calibrated scales normal to the incident beam and a well defined distance d = 129 mm.
b j a j b o a o arctan ] (1) d d Where a o and b o are the readings on the scales a and b respectively before the measurements and a j and b j the
= [arctan
1 2
readings on the scales at the measurement. The use of He -Ne Laser resulted in a sharp image on the light beam simultaneously on both the scales. To ensure good reproducibility and high accuracy, the BN crucible has to be completely filled with the melt at each measurement. The excess of the melt could rise in the capillary in the crucible lid and had thus no influence on the results. At the bottom of the crucible is a Pt-Pt/13%Rh thermocouple that measured the temperature. V. WORKING PRINCIPLES OF THE APPARATUS When a metallic specimen is placed in the centre of a rotating magnet field, an eddy current is induced in the specimen. By the mutual interaction between the external magnetic field and an additional magnetic field induced by the eddy current, a mechanical moment, inversely proportional to the electrical resistivity of the specimen. In case of liquid specimen, the rotational moment of the liquid in its container causes rotation of the container through the viscosity of the liquids and the rotating moment M is given by
M =
Where = the angular velocity of the rotating magnetic field, H = the external rotating magnetic field strength, = the viscosity of the specimen, = electrical conductivity of the specimen, L = length of the specimen and R = radius of the specimen respectively. By proper control of the dimensions of the specimen and the field strength H, the second term of the equation (2) becomes negligibly small. In case of metals of unknown , the rotating moment induced in the specimen can be balanced by the torsion of the strip, the lower end of which is attached to the container containing the specimen and the resistivity can be obtained by measuring the torsion angle as
. ..L.R 4 .H 2
192
1 . . 2 ..L.R 6 .H 4 ...(2)
M =
=C
H
2
LR 4 H 2 = K ...(3)
i2
as H ..( 4 )
and
=C
Where C =
4K = Apparatus constant, depends on the specimen dimensions, frequency of the rotating magnetic field LR 4 and magnetic field strength, = torsion angle and i = current intensity.
The exact derivation of the equations can be found elsewhere [17, 25-26] . VI. SAMPLE PREPARATION
Samples were prepared using high purity metals from the American Smelting and Refinery Company, USA and from Ventron Company, FRG. To remove the oxide layer (if any), the metals were further purified by melting the metals under vacuum and filtering the metal through quartz wool. The metals were weighed on an analytical balance and placed in a graphite crucible, which had exactly the same inner diameter as the BN- crucible used for the measurements. The graphite
33
crucible was sealed under a vacuum of 10-2 Pa in quartz; the metals were melted in a oven. The melted metals were kept for a week at 1273oK. After quenching in normal water, the specimens were removed from the graphite crucibles and grounded to the desire length of 20.5 mm. This procedure had to be followed since some of the metals were very brittle. VII. EXPERIMENTAL PROCEDURE The samples were put into the BN crucible which were connected to the torsion system and the whole arrangement was placed into the apparatus with the crucible at the centre of the alumina tube. The system was evacuated up to 10-2 Pa under argon atmosphere. The samples were then heated slowly under slow flow of argon gas. At different temperature 5-7 measurement were taken in the solid state. The sample was slowly heated (50oC/hour) up to about 20oK above the liquidus temperature and remained about 2 hours at that temperature. After a constant temperature was attained, the zero readings ( a o , bo ) of the torsion angles were taken. By changing the coil currents from 0.5A to 5.5A, up to ten measurements of the torsion angles ( a j , b j ) and of the corresponding current readings were taken. The arithmetic mean of the current readings of the three phases were calculated from
i1 + i2 + i3 (5 ) 3 2 Thus for each temperature 10 values of / i were calculated. Since these values showed slide drift, the arithmetic mean of i=
these data point was used to calculate the value of the conductivity of the given temperature. The temperature was then again raised to 20oK and waited till the temperature was again constant, the next measurements were taken. Since it was not possible to complete a series of measurements for a sample in one day, the samples were kept overnight in the molten state in the apparatus. The results of the measurements on the consecutive day agreed well with one another. In this way about 10-12 measurements from liquidus temperature up to 200oK above the liquidus temperature was carried out. Each time the samples were heated to the desired temperature and after a constant temperature was attained, the readings were taken. To cheque the reproducibility of the results, experiments were also carried out when the samples were cooled. The results on both the heating and cooling were in very good agreement. These samples were used to utilize to measure the value of the electric conductivity of the metals. VIII. EXPERIMENTAL RESULTS & DISCUSSION 1. Calibration of the apparatus Constant C : For the evaluation of the results, the apparatus has been calibrated using sample of pure magnesium, gallium and lithium. The results are tabulated in table-1.
Metals used Gallium (Ga) Magnesium (Mg) Lithium (Li)
/i2
0
(973 K
1.14 A 1.04 A
24.75 x1 cm-1
cm
C 3.54 x10 A
4 2
Mean
3.47 x10 4 A 2 .496 x10 4 3 27.72 x10 cm-1 cm-1

cm-1
cm
0.218 131.90 x1 3.48 x10 4 A 2

cm
This value of the apparatus constant C was calculated using equation (4). The results of these values are well fit with the results of Busch et al [27], Cusack et al [28] and Roll et al [29] respectively. 2. Experimental Date : The experimental electrical resistivity measurements were carried out on nine pure metals over the temperature range from liquidus temperature up to 1000K. The data obtained from heating and cooling were in good agreement. The experimental points about 12-17 data for each metals were taken. The results for few temperatures are listed in Tables 2 ~ 10.
Table 2: Mercury(Hg) of M.P.-38.9 oC*
34
Measurement Temperature ( 0C) -19 -11 0 13 44 100 145 200 240 275 320 340 362 400
Measurement Resistivity x 106 94.8 95.2 96.4 97.5 100 105.7 110.7 120.3 128.8 135.2 146.7 152.2 155.7 157.3
M.P.=Melting Point, * = Data from [12 ] Table 3: Tin (Sn) of M.P. 231.9 oC
Measurement Temperature ( 0C) 265 295 325 348 379 440 485 563 600 668 718 783 847 Measurement Resistivity x 106 48.6 49.5 50.1 50.9 51.5 52.2 52.9 53.8 54.6 55.6 56.2 57 57.6
Table 4: Bismuth (Bi) of M.P. 231.9 oC Measurement Measurement Temperature ( 0C) Resistivity x 106 279 128 325 129.7 375 133 400 134.8 440 136.8 471 138.1 526 140.9 550 142.8 590 144.9 639 146.8 680 148.9
35
709 752 795
150.9 152.1 155.8
Table 5: Cadmium(Cd) of M.P 320.9 oC Measurement Measurement Resistivity x 106 Temperature ( 0C) 351 32.01 392 32.18 419 32.23 457 32.32 494 32.48 528 32.56 550 32.67 574 32.72 633 32.92 650 33.98 700 33.18 740 33.26 762 33.29 Table 6: Lead (Pb) of M.P. 327 oC Measurement Measurement Resistivity x 106 Temperature ( 0C) 348 93.6 375 94.6 420 96.7 472 99.6 517 102.5 550 103.5 578 104.6 602 106.8 641 109.1 682 110.6 731 112.9 776 114.8 836 119.6 900 121.8
Table 7: Zinc (Zn) of M.P. 419.39 oC

Measurement Temperature ( 0C) 445 484 512 539 570 627 669 695 719 737 749 Measurement Resistivity x 106 37.2 37 36.8 36.6 36.3 35.6 35.1 34.8 34.6 34.4 34.2
36
763 801 852
33.9 33.6 33
Table 8: Antimony(Sb) of M.P 630 oC Measurement Resistivity x 106 Measurement Temperature ( 0C) 658 115.5 708 116.1 746 116.6 808 118.3 861 119.2 913 120 945 120.7 970 121.2 990 121.9 1019 122.6 1049 123.3 1091 124.2 1118 124.5 1152 125.3 Table 9: Aluminium (Al) of M.P.658.5 oC Measurement Measurement Resistivity x 106 Temperature ( 0C) 686 26 715 26.4 745 26.8 774 27.4 806 27.9 831 28.1 853 28.6 871 28.9 900 29.4 930 29.8 980 30.4 1023 30.9 1060 31.5 1100 31.9 Table 10: Silverc (Ag) of M.P. 960 oC Measurement Temperature ( 0C) 980 1001 1025 1047 1070 1096 1121 1143 1161 1173 1188 1199 Measurement Resistivity x 106 17.5 17.9 18.3 18.8 19.1 19.6 20 20.4 20.7 20.9 21.2 21.5
37
1206 1210
21.7 21.9
The calculated values of the resistivity varied for Aluminum from 26.0 x 106 31.9.x 106 between temperature range 6861100 oC; Mercury from 94.8 x 106 - 157.3 x 106 between temperature -19 400 oC; Tin from 48.6 x 106 - 57.6 x 106 between temperature 265 -847 oC; Bismuth from 128 x 106 155.8 x 106 between temperature 279-795 oC; Cadmium 32.01 x 106 33.29 x 106 between temperature351-762 oC; Lead from 93.6 x 106 121.8 x 106 between temperature348-900 oC; Zinc from 37.2 x 106 - 33.00 x 106 between temperature 445-852 oC; Antimony from 115.5 x 106 - 125.3 x 106 between temperature 658-1152 oC; and Silver from 17.5 x 106 - 21.9 x 106 between temperature980 - 1210 oC. The experimental data were compared with the data from the literatures. It was found 98% of the data were well fit with the literature. In some metal a slide variation of 2-3% were observed, this may due to impurities in the metal. IX. REFERENCES [01] [02] [03] [04] [05] [06] [07] [08] [09] [10] [11] [12] [13] [14] [15] [16] [17] [18] [19] [20] [21] [22] [23] [24] [25] [26] [27] [28] [29] de la Rive L, Arch. des. Sci. Phys. (Geneva) 17, 18, 362, 1863 Vincentini G and Omodei D, Atti. Acc. Soc. Torino, 20, 1884 Northrup E F, Trans. Amer. Elektrochem. Soc. 25, 373, 1914 Tsutsumi H. et al, Sci. Rep. 7, 93, 1918 Vassura et al, Neuvo. Cimento, 1890 Mller et al , E. T. Z, 13, 1892 Hackspill et al, Compt. rend. 151, 305, 1910 Bridgman et al, Proc. Amer. Acad. 56, 61, 1921 Bernini et al, Phys. Zeits. 5, 406, 1904 & 6, 74, 1905 Matiur R Mian et al, Bangladesh J.Sci Ind. Res., XXXI, 1, 1996 Scala E and Robertson W D , J Metals, N Y, 5, 1141-47, 1953 Matuyama Y, Science Report, Tohuku University, 16, 447, 1927 Nyberg D W and Burges R E, Can J Phys, 40, 1174, 1962 Tomlinson J L and Lichter B D, Trans.Met.Soc, AMIE, 245, 2261, 1969 Miller E, Paces J and Komarek K L, Trans. Met. Soc. AIME, 230, 1557, 1964 Schurnberger W, Thurn H and Krebs H, Ber. Bunsen Ges., 80, 730, 1976. Brawenbeck W, Z Phys, 73, 312 334, 1932 Grube G. and Spiedel P Z, Elektrochem., 46, 233 242, 1940 Roll A, Felger H. and Motz H , Z. Metallkunde, 47, 707 713, 1956. Regel, Ph.D Dissertation, University of Leningrad, U.S.S.R., 1957 Loffe A F and Regel A R , Progr. Semiconductor, 4, 237 291, 1960 Samarin A M , J. Iron St. Inst., 200, 95 101, 1962 Takeuchi S. and Endo H., Trans. Japan Inst. Metals, 3, 35 41, 1962. Liquid Metals, Chemistry and Physics by Sylvan, Z Beer, Maercel Dekker, Inc, New York, 1972 Eder O J, Kunsch B, Lacom W and Neumann W, Ost. Studienges f. Atomenergie, SGAE Ber No. 2194, Seibersdorf, Austria, 1973 Eder O J, Kunsch B, Lacom W and Neumann W, Ost. Studienges f. Atomenergie, SGAE Ber No. 2379, Seibersdorf, Austria, 1974 Busch G and Tieche Y, Phys Kondens. Mat., 1, 402, 1967 Cusack N E and Kendall P, Proc. Phys. Soc., 75, 309-311, 1960 Roll A and Motz H, Z. Metall, 48, 272-280, 1957.
38
I. Introduction
Mobile cellular network is growing rapidly and simultaneously the user of the cellular network is growing. As a result the number of new calls and handover calls in limited cells are increasing in a tremendous rate. Handling handover calls and new calls became a big challenge for cellular network. There are different methods for call admission for new calls and handover calls given in [1-2]. Two types of call admission, cell based call admission and tree based call admission are compared in this paper. Figure 1 shows a structure of a mobile cellular network. The fixed network is attached with two different switches. Each of the switches is covering 7 cells. Tree base call admission is shown under switch A and cell based call admission is shown under switch B. A mobile phone subscriber (also called mobile host) use one of this cell, when he/she receives a new call or while he/she travels among the cells. In this paper we will try to compare whether a cell based call admission or tree based call admission is more efficient for a MH (mobile host). Here is considered as new call arrival rate/cell, h is considered as handover rate/cell, t is considered as call termination rate for new call and h is termination rate for handover calls. Here number of user is unlimited. Details of call admissions are explained in section II. Network traffics are analyzed using state transition diagram in section III. In section IV results of the analysis are discussed. Finally section V gives concluding remarks.
Call Admission Control

TREE BASED CALL ADMISSION CONTROL In case of tree based call admission control a new call or a handover call is admitted to a group of 7 cells which is shown in Figure 1. So, new call arrival rate for the group of 7 cells is 7. Handover rate for the tree is 3h as there are 18 outside edge for the tree and handover rate to an edge of the cell is h/6 (18 X h/6 = 3h) which is also mentioned in [1]. Call termination rate is t for new call arrival and h for handover calls. Here 7C is total number of channels for the tree where N number of channels used for both new calls and handover calls. 7C-N channels are reserved for handling only handover calls.
Figure 1. Mobile Cellular Network
7C 3h N 7 0
Figure 2. Blocked diagram for tree based call admission
handover
new + handover
CELL BASED CALL ADMISSION CONTROL As the new call in admitted cell wise the new call arrival rate for the cell based technique is and call termination rate for new call is t., handover rate is h and termination for handover is h. Here C is total number of channels in a cell where m channels are both for new and handover calls and C-m is reserved for only handover calls.
1. 2.
Department of Computer Science and Enginering, IBAIS Universit. e-mail: mmrcse93@gmail.com. Warid Telecom, Gulshan, Dhaka. e-mail: sharrukhzaman@yahoo.com
Modeling a Medical Diagnosis System using Bayesian Belief Network
C h m 0
Figure 3. Blocked diagram for cell based call admission
7+3h 0 t 1 2t 7+3h 7+3h 7+3h 2 3t Nt N Nt+h 3h N+1 3h N+2 3h 3h 7C Nt+(7C-N)h
handover
new + handover
Nt+2h
Figure 4. State transition diagram for tree based call admission control
+h 0 t 1 2t +h 2 3t mt +h +h m mt+h h m+1 h m+2 h h C mt+(C-m)h
mt+2h
Figure 5. State transition diagram for cell based call admission control
III. Traffic Analysis

State transition diagram [3] for tree based call admission is shown in figure 4. And for cell based call admission control shown in figure 5. By analyzing the state transition diagram in figure 4 and using cut equation we find following derived equations: Blocking probability for new call arrival in a tree based technique is
7. Pbn( ) t N i= 0 7. 3. h . 1 i! t
i
3. h
.1 N! .1 . N! 7. C N j= 1 j ( i. h ) i= 1
j ( 3. h )
7. t
3. h
7. t
3.h
. . 1 . ( 3 h) N ! 7.C N i= 1
7 .C N
Probability of handover failure in a tree based technique is
( i. h ) Phf( ) N i= 0 7. C N j= 1 j ( i. h )
7.
i 3. h . 1 i! t
7. t
3. h
.1 . N!
( 3. h )
i= 1 By analyzing the state transition diagram in figure 5 which is also used in [4-10] and using cut equation following equations are derived Blocking probability for new call arrival in a cell based technique is
Pbn( ) t m i= 0 h .1 i! t
i
. 1 m! C m j= 1 j ( i. h ) i= 1 h
j
. 1 . m!
Probability of handover failure in a cell based technique is
31
Md. Mizanur Rahman et al.
h . 1 . m! C m i= 1
( i. h ) Phf( ) m i= 0 C m j= 1 j ( i. h ) i= 1
h .1 i! t
. 1 . m!
IV. RESULT AND DISCUSSION
Blocking probability of new arrival and handover failure of both cell and tree based admission is plotted against new call arrival rate taking number of allocated channel of new arrival call and guard channel of handover call as parameters shown in figure 6-9. Here handover arrival rate h=0.2 calls/min, handover termination rate h =1.2 calls/min, total termination rate t = 2.6 calls/min and total number of channels C=8 taken for cell based call admission technique. For tree based admission, number of channels for new call N=14 and average number of channels per cell C=3. Both blocking probability of new arrival call and handover failure increase with increment of call arrival rate at the same time both probability decreases with increment of allocated channel of respective traffic.
When large number of users shares a common string of channels usually show better performance than the situation when several group of channels are allocated for individual group. The situation is also visualized in two call admission schemes i.e. tree based admission shows better performance with less number of channels/cell.
Tree Based Call Admission Blocking Pr. of new call
0.1 0.01
1 .10 4 1 .10
1 .10
5
6
1 .10 7 1 .10
1 .10
1 .10
8
9
2 3 New call arrival rate (calls/min)
N=14, C=3 N=16, C=3 N=18, C=3
Figure 6. Blocking probability against new call arrival rate for tree based call admission
1 .10 1 .10 1 .10
6
Tree Based Call Admission
Pr. of Handover filure
1 .10 10 1 .10 11 1 .10 12 1 .10 13 1 .10 14 1 .10 15 1 .10 .10 16 1 17 1 .10
1.5
2 2.5 3 New call arrival rate (calls/min)
3.5
N=14, C=3 N=16, C=3 N=18, C=3
Figure 7. Probability of handover failure against new call arrival rate for tree based call admission
32
Cell Based Call Admission Blocking Pr. of new call

0.1 0.01 1 .10 1 .10 1 .10 1 .10 1 .10
3 4 5 6 7
m=5, C=8 m=6, C=8 m=7, C=8
Figure 8. Blocking probability against new call arrival rate for cell based call admission
1 .10
3
Cell Based Call Admission
Pr. of Handover filure
1 .10
1 .10
1 .10
1 .10
1 .10
m=5, C=8 m=6, C=8 m=7, C=8
Figure 9. Probability of handover failure against new call arrival rate for cell based call admission
V. Conclusion
All the graphs of previous section yield logical results. Here traffic is modeled using one dimensional Markovian chain but result would be more accurate if two dimensional chain (where any probability sate (i,j) ; i=number of new call and j=number of handover call) were used. Inclusion of queue with guard channel could further enhance carried traffic of handover part.
VI. References
[01] [02] [03] [04] [05] [06] [07] [08] [09] K. Wang and L. Lee, Design and Analysis of QoS Support Frequent Handover Schemes in Microcellular ATM Networks, IEEE Transaction on Vehicular Technology, Vol. 50, No. 4, pp942-953 July 2001 Y. Fang and Y. Zhang, Call Admission Control Scheme and Performance Analysis in Wireless Mobile Network , IEEE Transaction on Vehicular Technology, Vol.51, No.2, pp. 371-382, March2002 J.F.C. Kingman, Markov population processes. J. Appl. Prob., Vol. 6 (1969), 1-18, 1969 G.V Tsoulos, Expermental and theoretical capacity analysis of space-division multiple access (SDMA) with adaptive antennas IEEE proccommunications, vol.146, No.5, pp. 307-311, Oct1999 V. B. Iverson, Teletraffic Engineering Handbook, Technical University of Denmark, November 2001 M. I. Islam, M. Q. Maula, and L. J. Rozario, Analysis of Two Dimensional Limited Source and Mixed Traffic Model for a BTS of a Small Mobile Cellular Network Proceedings of ICCIT 2001, pp190-196 M. S. Hossain and M. I. Islam A Proposed 2-D Queueing Model of PCT-I Traffic Proceedings of ICCIT 2003, vol. I, pp114-118. M. S. Hossain and M. I. Islam, A Theoretical Analysis of SDMA Traffic of Limited User Network for Duplicate at Last Case, Proceedings of ICCIT 2004, pp 728-732. M. S. Hossain and M. I. Islam, Analysis of SDMA PCT-II Trafficfor Duplicate First Case -A Theoretical Approach, Proceedings of 2005 International Conference on Wireless Communications Networking and Mobile Computing,Vo. I, pp 423-426, Wuhan University, Wuhan , China
33
Md. Mizanur Rahman et al.
[10] [11]
Kwan L. Yeung and Sanjib Nanda, Channel Management in Micro/Macro Cellular Radio Systems IEEE Transaction on Vehicular Technology,Vo.45, No.4, pp.14-23, Nov1996 S. A. El-Dolil, W.C. Wong, and R. Steele. Teletraffic performance of highway microcells with overlay macrocell. IEEE Journal on Selected Areas in Communications, 7(1):71,78, January 1989.

Md. Mizanur Rahman1 and Md. Sharrukh Zaman2
ABSTRACT
Bayesian belief networks are useful tools for coherent representation of uncertain knowledge. They are widely used in medical diagnostic reasoning. There are many expert systems to help physicians diagnosing different classes of diseases. In this paper we propose a system using belief networks to do the primary diagnosis of common diseases. Here a detailed study of physicians working procedure has been performed and it is modeled in a belief network. So far as we know, this is the first of its kind. This will be a very much helpful tool for primary diagnosis.
Keywords: Bayes Rule, Belief Network, Medical Diagnosis. I. BACKGROUND Many AI techniques have been used in medical diagnostic reasoning. Among them Bayesian belief network remains in a higher position because it allows human to apply expertise and prior knowledge directly into the computations. Bayes rule of computing conditional probability is the foundation of belief networks. A belief network is a directed acyclic graph whose nodes correspond to random variables. Links denote direct dependencies and the conditional probabilities for directly dependent variables are part of the network. A generic singly connected network is shown in the figure1. Here Ex+ is the casual support and Ex- is the evidential support for X, which combines to form E. P(X|E) is calculated using backward-chaining algorithm derived by Russell and Norvig [3]. For convenience, derivation of the algorithm is given below. Our aim is to compute P(X|E) given that X is not in E. To mention again, E is divided into two parts - one is Ex+, the casual support for X, i.e. the evidence variables that are its parents or connected via the parents of X. The other part is Ex- - the evidential support for X, i.e. the evidence variables that are children of X or connected via children of X. Hence we have, P (X|E) = P (X| Ex-, Ex+) 34
Ex-
= P (Ex-|X, Ex+) P (X|Ex+)/P (Ex-|Ex+), Keeping as fixed background evidence.
U1
EX+
Um
Znj Ym
EX
Znj Ym
Figure 1: Generic singly connected network.
Structural Analysis of Unicode Based Keyboard: Bengali Perspective From An Operational Standpoint
Using conditional independence and normalizing constant, P (X|E) = P (Ex-|X) P (X|Ex+). To calculate P (X|Ex+), let U be the vector of U1,, Um, and u be an assignment of values to them. Then we have, P (X|Ex+) = u P (X|u, Ex+) P (u|Ex+) = u P (X|u) i P (ui|Ex+), using d-separation and conjunction of independent variables. Therefore, P(X|Ex+) = u P(X|u)i P(ui|EUi\X), partitioning Ex+ into EU1\X,, EUm\X and using the fact that EUi\X d-separates Ui from all other evidence in Ex+. Now to calculate P (Ex-|X), we find that the evidence in each Yi box is conditionally independent of the others given X. Therefore, P (Ex-|X) = i P (EYi\X|X) = i yi zi P (EYi\X|X, yi , zi) P (yi, zi|X), averaging over Yi and zi, where Zi is the parents of Yi other than X and zi is - an assignment of values to the parents. = iyiziP(EYi-|X, yi , zi)P(EYi+\X|X)P(yi, zi|X), dividing EYi\X into EYi- and EYi+\X. = i yiP(EYi-|yi)ziP(EYi+\X|zi)P(yi, zi|X), applying conditional independence. =iyiP(EYi|yi)zi(P(zi|EYi+\X)P(EYi+\X)/P(zi)) P(yi,zi|X), applying Bayes rule. = i yiP(EYi-|yi)zi(P(zi| EYi+\X)P(EYi+\X)/P(zi)) P(yi|X, zi) P(zi|X), rewriting the conjunction Yi, zi. = i yiP(EYi-|yi)zi i P(zi| EYi+\X) P(yi|X, zi), applying conditional independence and replacing P(EYi\+X) by normalizing constant i Finally, multiplying the Zijs as they are independent of each other and combining the is into one big normalizing constant , we have P(Ex-|X ) = i yi P(EYi-|yi)ziP(yi|X, zi)j P(zij| EZij\Yi). Now we are ready to calculate P(X|E) = P(Ex-|X) P(X|Ex+). II. STUDY AND FINDINGS After getting the patients identity, physicians note the age, sex, religion, locality, occupation, and economic condition of a patient. These parameters are essential because probability of many diseases depends on them. Then details about the complaints of the patient are asked. For example when a patient complaints of fever or sweating etc. Answers to these queries help the physician to differentiate the complaints into specific group of diseases. We have selected diseases with a common symptom, fever, for primary study. Most common of them are non-specific viral, malaria, visceral leishmaniasis, enteric fevers, rheumatic fever, tuberculosis, pneumonia, urinary tract infection, meningitis, dengue and tonsillitis. Then fifteen characteristics to distinguish these diseases were identified. They are percentage among fever diseases, region, age, sex, and economic condition
dependency, magnitude, onset, duration and time dependency of temperature, presence of body pain, shivering or sweating, association of burning sensation while passing urine, soar throat or cough. To determine the network topology, we grouped the diseases characteristics into two- influence and symptom. First five of the above lists fall in the influence category and the rest fall into symptom category. Then the network for a particular disease can be expressed as in figure-2.
I1
I2
Im
D
S1 S2 Sn
Figure 2: Network for a particular disease. Here D stands for the disease node, I1, I2,, Im for influence nodes and S1, S2,, Sn for symptom nodes. The assignment of probabilities P (D|Ii) and P (Sj|D) can be done using the experience of practicing physicians and statistical data provided by some hospitals. Medial textbooks also helped in this matter. We applied Bayes rule to calculate some probabilities not directly found from collected data. In this rule, we devised a simulator using some test data to determine the performance of our model. It came out that four diseases are successfully diagnosed more than 90% of the time. Success rate for two other diseases are around 70%. Less frequent three diseases are correctly diagnosed around 50% of the time. As our model used probability theory, correctness of diagnosis of two-rate disease was significant in percentage. We show the summary of 100,000 simulation runs for two diseases below. The summary shows the number of times each position (by probability value) obtained by the simulated disease in the diagnosis phase. Also average position and standard deviation are calculated shown in table-1.
Summary of 100,000 simulation runs for the disease influenza Number of times each position obtained Position 1: 96911 Position 2: 2533 37
Mohammad Shafkat Amin et al.
Position 3: 412 Position 4: 115 Position 5: 22 Position 6: 4 Position 7: 3 Percentage of Success: 96.911 Average position: 1.03828 Standard deviation: 0.23726 Summary of 100,000 simulation runs for the disease malaria Number of times each position obtained Position 1: 91624 Position 2: 5206 Position 3: 1839 Position 4: 590 Position 5: 331 Position 6: 243 Position 7: 102 Position 8: 58 Position 9: 7 Percentage of Success: 91.624 Average position: 1.14267 Standard deviation: 0.58454 Table 1: Summary of simulation runs for influenza and malaria. III. CONCLUSION There is a high possibility of this field in our country and other poor countries like ours, since physicians are inadequate in number and not available everywhere. If we can provide software in at least every Thana health complex in our country, it can bring a revolutionary change in the service of health sector. Besides, such software can be very much helpful for novice doctors by giving them a guideline about identifying the disease of their patients from primary complaints. It can also be used as a time saving tool for experienced and specialist physicians.
of Internal Medicine Vol. 108; No. 1, pp. 80-87. January 1988.
IV. REFERENCE [01] Long, W. Medical Diagnosis Using a Probabilistic Casual Network. Applied Artificial Intelligence, 3: 367-368, 1989. [02] [2] Macleod, J., Editor. Davidsons Principles and Practice of Medicine. English Language Book Society / Churchill Livingstone. [03] [3] Russell, S.J. and Norvig, P., 1995. Artificial Intelligence: A modern Approach. Prentice Hall International, Inc. [04] [4] Szolovits, P., Patil, R.S. and Schwartz, W.B. Artificial Intelligence in Medical Diagnosis. Annals 38
Structural Analysis of Unicode Based Keyboard: Bengali Perspective From An Operational Standpoint
Structural Analysis of Unicode Based Keyboard: Bengali Perspective from an Operational Standpoint
Mohammad Shafkat Amin1, Fahima Amin Bhuiyan2 and Mohammad Shyiq Amin3
ABSTRACT
This study investigates the Unicode specifications and the possibilities for their implementation in Bengali Script Language. Unicode provides a unique number for every character, irrespective of the platform, program and the language. These characters cover principal written languages and symbol system of the world. The study explores an efficient way of incorporating Unicode based Bengali keyboard in contemporary operating systems. This paper systematically investigates an in depth analysis of Unicode based Bengali keyboards and its implementation aspects unfolding the details of supports provided by the operating system as well as explaining the necessary technical prospects of open type fonts. Keywords: Unicode, Glyph, lparam, Codepage, OTF.
I. INTRODUCTION Unicode provides a unique number for every character irrespective of the platform, program and the language. Fundamentally, computers just deal with numbers. They store letters and other characters by assigning a number for each one. Before Unicode was invented, there were hundreds of different encoding systems for assigning these numbers. No single encoding system could contain enough characters: for example the European Union alone requires several different encoding to cover all its languages. Even for a single language like English no single encoding was adequate for all the letters, punctuations, and technical systems in common use [1]. These encoding systems also conflict with one another, that is, two encodings can use the same number for two different characters, or use different number for the same character. Any given computer (specially servers) needs to support many different encodings, yet whenever data is passed between different encodings or platforms, the data always runs the risk of corruption [2]. This paper proposes an efficient way of implementing Unicode based Bengali keyboard in modern operating systems. This approach facilitates the linguistic computing in the local domain and enables the support of multi-lingual computing as well. II. THE UNICODE STANDARD The Unicode worldwide character standard is a character coding system designed to support the interchange, processing and display of the written texts of the diverse languages of the modern world. In addition, it supports classical and historical texts of many written languages [1, 2]. In its first 256 characters, it includes the repertoire of characters from numerous other corporate, national and international standards as well. The Principles of Unicode standards A team of computer professionals linguists and scholars created the Unicode to become a worldwide standard, one easily used for text encoding everywhere. To the end, it follows a set of fundamental principles. The Unicode standard is simple and consistent. It uses fixed width 16 bit character codes, and does not depend on states or modes for encoding special characters. The Unicode standard incorporates the character sets of many existing standards, for example, it includes Latin-I, character set as its first 256 characters [2]. It includes repertoire of characters from numerous other corporate, national and international standards as well. The Unicode standard uses Han unification to consolidate Chinese, Korean and Japanese ideographs. The Unicode standard allows for character composition in creating marked characters. It encodes each character and diacritic or vowel mark separately, and allows the characters to be combined to create a marked character. It provides single codes for marked characters when necessary to comply with preexisting character standard. III. UNICODE CAPABILITIES 1. Role of Unicode specification in Internationalization
39
Upgrading Trust Factor Evaluation in AODV Protocol for MANET
Unicode is the new foundation for this process of internationalization. Older code pages were difficult to use and had inconsistent definitions for characters. Internationalizing different codes while using the same code base is complex, since different character sets with different architectures for different markets, was needed to be supported. But modern business requirements are ever stronger; programs have to handle characters from a wide variety of languages at the same time. The EU alone requires several different older character sets to cover all its languages. Mixing older character sets together is a nightmare, since all data has to be tagged and mixing data from different sources is nearly impossible to do reliably [3]. With Unicode, a single internationalization process can produce code that handles the requirements of all the world markets at the same time. Since Unicode has a single definition for each character, data corruption problems that plague mixed code set programs, are easily avoided. Since it handles characters for all the world markets in a uniform way, it avoids the complexities of different character code architectures. All of the modern operating systems, from PCs to mainframes, support Unicode now, or are actively developing support for it. 2. Unicode specification Vs using classical character sets for application programs Different character sets have very different architectures. In many, even simply detecting which bytes from a character is a complex, contextually dependent process. That means either having ultiple versions of the program code for different markets, or making the program code is much, much more complicated. Both of these choices involve development, testing, maintenance, and support problems. These make the non-US versions of programs more expansive and delay their introduction, causing significant loss of revenue. 3. Using classical character sets for databases Classical character sets only handle a few languages at a time. Mixing languages was very difficult or impossible. In todays markets, mixing data from many sources all around the world, that strategy for products fails badly. The code for a simple letter like A will vary wildly between different sets, making searching, sorting and other operations very difficult. There is also the problem of tagging every piece of textual data with a character set, and corruption problems when mixing text from different character sets [4]. 4. Bengali in Unicode The problem for Indic languages is more complex compared to European languages. Unlike Latin-based scripts, here there is a very basic distinction between a sequence of characters and the rendering of these characters in a visually acceptable form. It should be noted that Unicode says little about how to actually display a sequence of abstract characters. Unicode attempts to list all alphabetic characters (as opposed to glyphs [5, 6], the visual form representing one or more characters) used in all the major languages of the world. Different groups of languages use different sets of characters, many characters are common to more than one language (e.g., Hindi and Sanskrit use the same character set; the French character set is a superset of the English one). Unicode tries to list all these characters. Bengali characters are numbered from 0980 to 09FF (Devanagari from 0900 to 097F). The advantage of Unicode is that Unicode Bengali characters are uniquely and unquestionably Bengali. Even though the question of how they should be displayed is not clear, the intended content of the file is uniquely defined as shown in figure 1.
Figure 1: Bengali Codepage in Unicode IV. BENGALI KEYBOARD IN WINDOWS All windows in a Windows program are created based on a window class, which is a template that defines the attributes of a window. Multiple windows can be created from a single window class. The window class must be registered with Windows using the RegisterClassEx() Win32 API function. After a window class has been registered, it can be used to create as many windows wanted. Window classes are represented by a data structure in the Win32 API called WNDCLASSEX, which defines the attributes of a window [7].
41 JournalofComputerScience,Volume1,Number1,June2007,ISSN:19946244
Muhammad Ruhul Amin Khandaker et al.
Windows communicates with application programs by sending it messages. A message has three pieces of information associated with it: A window A message identifier Message parameters The window associated with a message is the window to which the message is being sent [7]. The DispatchMessage(&msg); } This code essentially processes all messages at the application level and routes them to the appropriate window procedures. Code in each different window procedure is then responsible for taking action based on the messages they receive. Windows message processing system is depicted in figure 2. In order to programmatically incorporate a Unicode based keyboard in windows, a system wide message hook has to be established. A hook is a mechanism by which a function can intercept events (messages, mouse actions, keystrokes) before they reach an application. The function can act on events and, in some cases, modify or discard them. Functions that receive events are called filter functions and are classified according to the type of event they intercept. For example, a filter function might want to receive all keyboard or mouse events.
Figure 2: Windows Message Processing System. For message identifier is a number that specifies the message being sent. The Win32 API defines numeric constants that represent each message. The message parameters consist of two pieces of information that are entirely specific to the message being sent. These 32-bit values are called wParam and lParam, and their meaning is completely determined by the message being handled [7]. When a message is delivered to a program by Windows, it is processed in the WndProc() function. Although WndProc() is responsible for handling messages for a given window class, programs must still take on the task of routing messages to the appropriate window procedures. This is taken care of by a message loop, which must be placed in the heart of the WinMain() function: while (GetMessage(&msg, NULL, 0, 0)) { TranslateMessage(&msg); DispatchMessage(&msg); } Windows to call a filter function, the filter function must be installed that is, attachedto a Windows hook. Attaching one or more filter functions to a hook is known as setting a hook. If a hook has more than one filter function attached, Windows maintains a chain of filter functions. The most recently installed function is at the beginning of the chain, and the least recently installed function is at the end. This is depicted in figure 3. When a hook has one or more filter functions attached and an event occurs that triggers the hook, Windows calls the first filter function in the filter function chain. This action is known as calling the hook. To maintain and access filter functions, applications use the SetWindowsHookEx and the UnhookWindowsHookEx functions [7]. The following code snippet when executed establishes a windows hook.
Figure 3: The filter function chain in Windows [7]. HOOK=SetWindowsHookEx(WH_GETMESSAGE,(HOOKPROC)Key,hDLL,0); Here the WH_GETMESSAGE parameter installs a hook procedure that monitors messages posted to a message queue and the parameter Key is a pointer to the hook procedure. The zero in the last parameter specifies that, the Key parameter must point to a hook procedure in a dynamic-link library (DLL). Otherwise, this parameter can point to a hook procedure in the code associated with the current process.
JournalofComputerScience,Volume1,Number1,June2007,ISSN:19946244 42
The WM_CHAR message is posted to the window with the keyboard focus when a WM_KEYDOWN message is translated by the TranslateMessage function. The WM_CHAR message contains the character code of the key that was pressed. Here the Callback function Key actually intercepts the Windows messages and checks to see whether it is a WM_CHAR message and if so it performs the actual keyboard translation and sends back a Unicode character, which can be any of the valid Unicode character code i.e. 0x995. V. OPEN TYPE FONT SPECIFICATION The OpenType font format is an extension of the TrueType font format, adding support for PostScript font data. The OpenType font format addresses the following goals: broader multi-platform support better support for international character sets better protection for font data smaller file sizes to make font distribution more efficient broader support for advanced typographic control [8]. PostScript data included in OpenType fonts may be directly rasterized or converted to the TrueType outline format for rendering, depending on which rasterizers have been installed in the host operating system. OpenType fonts can include the OpenType Layout tables, which allow font creators to design better international and high-end typographic fonts. The OpenType Layout tables contain information on glyph substitution, glyph positioning, justification, and baseline positioning, enabling text-processing applications to improve text layout. As with TrueType fonts, OpenType fonts allow the handling of large glyph sets using Unicode encoding. Such encoding allows broad international support, as well as support for typographic glyph variants. Figure 4 represents the Glyph positioning in Bengali script development.
Figure 4: Glyph position in Bengali Alphabet [9] Whether TrueType or PostScript outlines are used in an OpenType font, the following tables are required for the font to function correctly. Tables 1,2,3,4 and 5 describe the names of these tables. Table 1: Required Table for OTF [9] Name Character to glyph mapping Font header Horizontal header Horizontal metrics Maximum profile Naming table OS/2 and Windows specific metrics post PostScript information Tag cmap head hhea hmtx maxp name OS/2 For OpenType fonts based on TrueType outlines, the following tables are used: Table 2: Tables Related to TrueType Outlines [9] Tag Name cvt Control Value Table fpgm Font program glyf Glyph data loca Index to location prep CVT Program
Muhammad Ruhul Amin Khandaker et al.
The PostScript font extensions define a new set of tables containing data specific to PostScript fonts that are used instead of the tables listed above. Table 3: Tables Related to PostScript Outlines [9] Tag Name CFF PostScript font program (compact font format) VORG Vertical Origin Table 4: Tables Related to Bitmap Glyphs [9] Tag EBDT EBLC EBSC Name Embedded bitmap data Embedded bitmap location data Embedded bitmap scaling data
OpenType fonts may also contain bitmaps of glyphs, in addition to outlines. Hand-tuned bitmaps are especially useful in OpenType fonts for representing complex glyphs at very small sizes. If a bitmap for a particular size is provided in a font, it will be used by the system instead of the outline when rendering the glyph. There are also several optional tables that support vertical layout as well as other advanced typographic functions: Table 5: Advanced Typographic Tables [9] Tag Name BASE Baseline data GDEF Glyph definition data GPOS Glyph positioning data GSUB Glyph substitution data JSTF Justification data
Thus all the information controlling the substitution and relative positioning of glyphs during glyph processing is contained within the font itself. This information is defined in OpenType Layout (OTL) features that are, in turn, associated with specific scripts and language systems. VI. CONCLUSION This paper investigates the Unicode specification and its possibilities in terms of Bengali script language and multilingual processing. This research paper explores the internal specification of Bengali Unicode, OTF and its support in modern operating systems for Bengali script processing. Specifically the support for Bengali script processing has been investigated in terms of Windows operating system and the windows message passing architecture has been explained and its implementation aspects have also been discovered. Nowadays, as the need for multi-lingual support in operating systems have become eminent, the importance of Unicode and its implementation aspects have gained greater attention. This paper simplifies the architecture of Unicode based Bengali language processing. VII. REFERENCES [01] [02] [03] [04] [05] [06] [07] The Unicode Consortium. The Unicode Standard, Version 4.0. Reading, MA, Addison-Wesley, 2003. 0-321-18578-1.z Memon, A.P. Study of Unicode specifications and their implementation in Arabic script languages by designing a multilingual Unicode editor.Multi Topic Conference, 2001. IEEE INMIC 2001. Technology for the 21st Century. Proceedings. IEEE International 28-30 Dec. 2001 Page(s):229 - 233 Bishop,Avery,F.,Brown,David,C., Meltzer, David, M. Supporting Multilanguage Text Layout and Complex scripts with Windows 2000. Microsoft System Journal. Becker, J.D. Multilingual word processing. Scientific American. 251, 1 (July 1984), 96-107. Becker, J.D. Arabic word processing. Communications of the ACM 30, 7 (July 1987), 600-610. Davis, M.; Collins, L. Unicode, IEEE International Conference on Systems, Man and Cybernetics, 1990. Conference Proceedings. 4-7 Nov. 1990 Page(s):499 - 504 Charles Petzold, Programming Windows, The Definitive guide to the Win32 API, 5th Edition, Microsoft Press, 2001.
[08] [09]
Mudawwar, M.F. Multicode: a truly multilingual approach to text encoding, Computer IEEE Journal, Volume 30, Issue 4, April 1997 Page(s):37 43. Microsoft, Typography, http://www.microsoft.com/typography.

Muhammad Ruhul Amin Khandaker1, Md. Javed Kaiser2, Jugal Krisna Das3 Md. Shahid Hossain1 and Khandaker Razia Sultana3
ABSTRACT
In an ad hoc network, users communicate with each other from a temporary network, without any form of centralized administration. Each node participating in the network acts both as host and a router and must therefore be willing to forward packets for other nodes. The existing wireless routing protocols do not accommodate enough security for MANET and are highly vulnerable to attacks as malicious nodes may enter and leave the immediate radio transmission range at random intervals. To reduce this vulnerability and ensure security some researchers introduced trust levels of each node to select the neighbors. In this paper we introduce a technique for upgrading the trust levels of each node. Key words: MANET, AODV, routing, secure neighbor, malicious node, trust factor/ level.
I. INTRODUCTION Ad hoc network does not rely on any stationary infrastructure. The concept behind these infrastructures less networks is the collaboration between its participating members, i.e., instead of making data transit through a fixed base station, nodes consequentially forward data packets from one to another until a destination node is finally reached. Typically, a packet may travel through a number of network points before arriving at its destination. The operation of the system depends on distributed cooperation among all nodes in the network and fairly needs to trust the intermediate nodes between sender and receiver and the neighbors. Because of the improvised nature of ad-hoc networks, routes are built dynamically as and when nodes are regrouping (Discovery). Hence, ad hoc networks are more responsive to topology changing than any wired networks. Consequently, routing protocols for ad-hoc networks should be able to cope with link breakages and make sure that the network wont collapse as nodes are moving or shutting down. This paper describes secure neighbor detection for mobile ad hoc networks. In particular, we employ a trust-level based technique to find the nodes which might be neighbors for the ad hoc on-demand distance vector (AODV) routing protocol, a widely adopted ad hoc routing protocol. AODV is a reactive and stateless routing protocol that establishes routes only as desired by the source node. AODV is vulnerable to various kinds of attacks [15]. This paper analyzes some of the vulnerabilities, specifically discussing attacks against AODV that related to neighbors. We propose a solution based on the trust level to upgrade trust level as a malicious node may enter in network any time and ensue it cannot be chosen as the neighbor. The remainder of this paper is organized as follows: Section describes the secure neighbor in AODV. Section describes relative work on AODV security and intrusion detection on ad hoc networks. Section describes the upgrading technique of trust level in AODV. And finally in section we summarize the work II. SECURE NEIGHBOR DETECTION IN AODV Differentiating between a node and malicious in ad hoc networking environment is a challenging task. Malicious nodes may behave maliciously only erratically, further complicating their detection. A node that sends out false routing information could be the one that has been compromised, or merely one that has a temporarily stale routing table due to volatile physical conditions. Dynamic topologies make it difficult to obtain a global view of the network and any approximation can become quickly outdated. Therefore, secured neighbor detection is the first task for secured routing in AODV. In a network a node initiates path discovery process while it needs to communicate with another node if it does not have sufficient information. This process is accomplished by broadcasting a route request (RREQ) packet to its neighbor. After receiving a RREQ packet
A Novel Segmented Display for Arabic Numerals
a node may reply back by forwarding RREP packet or rebroadcast RREQ to its neighbors. Before rebroadcast it will increase the hop field and remaining same the destination sequence number. The previous one is done if the receiving node is destination and the next one if it is an intermediate node. For a secured neighbor detection we propose to add several security modules with the existing AODV. In AODV protocol, a secure node wishing to communicate with a destination node, first broadcasts a RREQ packet to its neighbors. Upon receipt, the destination node reply RREP packet to the source. Each node maintains only the hop information to reach to destination. The route selection criteria of AODV based upon hop count and destination sequence number. Hop count determines how short the route is, and the sequence number of the destination speaks about the freshness of the route information. Therefore, the route selection metric is clearly independent of the security level of the application and trust factor of the participating nodes. Each node maintains a local database of its neighbors, where trust factor is dynamically updated 3]. Trust factor of each node calculated from the trust level is stored in the local database. Trust level of any node is defined from an integer value. Trust value is calculated from the activities of a node when routing occurs. Every node dynamically upgrades its trust level upon observing its neighbor. This protocol allows both the initiator and target to verify that both are within their maximum transmission range. It is a simple three-round mutual authentication protocol. In 1st round, initiator sends a neighbor solicitation packet by. The target sends the reply by a solicitation packet after receiving the packet. In last round, the initiator sends neighbor verification, including broadcast authentication of a timestamp and the link from the source to destination. Now the total delay between these two subsequent messages is found by = t1 t0. The distance between them (with respect to initiator) is bounded by: d <= /2 * c, where c is the speed of light in vacuum. Thus, the initiator can check that the other party is within its maximum transmission range. The process of secure neighbor detection is performed off line. Each node maintains a local database for maintaining a buffer with trust level of neighbor node. The following figure 1 shows the format of local database (with possible values): Target Packet ID Forwarded Unaltered Node <IP, Broadcast ID> A A, 0001 1 0 Figure 1: Format of the local database of each initiator node III. RELATED WORKS Yan et al. [16] proposed a trust model for secure routing evaluation in MANET. The authors defined a large trust evaluation matrix based on statistic data collected during the network communication. The statistic fields try to include different affective factors of the evaluation, such as pre-existing relationship among the nodes. A linear function is proposed to link these statistic fields together to compute the trust value about a certain node or nodes. However, no boundary evaluation value is defined in their approach. So it is difficult to define a threshold trust value for on-going tasks. Virendra et al. [17] proposed a pair-wise trust evaluation scheme in MANETs. To evaluate the trustworthiness of a target node, a node implements some self evaluation on the target node while also considering other nodes trust on the same target node. All trusts are evaluated via node monitoring on data delivery in the network. For computing self evaluation a traffic statistic function is mentioned, but not explicitly presented. To combine the self evaluation and others trust, a relationship equation is defined. In the equation, self evaluation and others trust are weighted with factor a1 and a2 respectively (a1+a2=1). The limitation of such relationship equation is that all different direct experiences are adjusted with one weight factor of same value. It is evaluated two types of trust between a trustor node and a trustee node: direct trust and recommendation trust. Direct trust is a kind of credential gained by a trustor node through its direct experience upon the trustee node. Recom-mendation trust is the credential gained by a trustor node from a third node or nodes recommendation on the trustee node. Jugal Krishna Das, Shareeful Islam, Abu Raihan Mostofa Kamal, et al. 3] proposed a notation for calculating the trust factor as T [ A , B, t ] = x. It implies that the trust factor of a node A at time t is x measured by node B. Therefore it is a relative measure as T [A, B, t ] T [ A , B, t ] is not a necessary condition. Trust factor of a node is determined by calculating the misbehaving of any node.
JournalofComputerScience,Volume1,Number1,June2007,ISSN19946244 45
(Y/N)
(Y/N)
Khondaker Abdullah-Al-Mamun et al
IV. A NEW MODEL FOR TRUST LEVEL MODIFICATION A predefined time period tm is set up and after the tm time all entry of the local table will be deleted. A predefined threshold value Pthreshold is the determination parameter to detect a node as malicious. Our objective is to count the successful forwards by the target node. It can be easily computed by simply performing logical AND operation of the forward and reverse field of figure 2. Then the total number of 1s generates the desired successful packet forwards. Target Node Packet ID <IP, Broadcast ID> A A A A A A A A A, 0001 A, 0002 A, 0003 A, 0004 A, 0005 A, 0006 A, 0007 A, 0008 0 1 1 1 1 1 1 1 0 0 1 0 1 1 1 1 Forwarded (Y/N) Unaltered (Y/N)
Figure 2: Local database with assumed value Suppose a snapshot for a period of tm is generated by a node. Now by AND operation of the last two fields the resultant transfer string becomes 00101111. So the number of the successful packet forwards Psuccess1 = 5 and the success factor ratio is SFRA = Psuccess1 / Ptotal = 5/8 =62.57%. Now D is calculated for a node A by subtracting the Pthreshold from the new calculated success factor ratio as D = SFRA - Pthreshold If D 0 then the target node is detected as malicious. As each node maintains a local database of its neighbors with corresponding trust factor. So, the initiator updates the local database setting the trust factor of target node 1. The use of SFR1 instead of absolute number of successful packet forward Psuccess has been carefully chosen. Upgrade of trust factor mechanism is slightly different from that of degrade mechanism. But it uses almost the same algorithm for building the transfer string. Upgrade mechanism has two predefined values, tt and Pthreshold which is necessarily equal values of tm and Pthreshold respectively of the degrading process. The upgrading ratio will not be same as the degrading ratio. The value Psuccess2 is calculated by summing up the number of consecutive 1s from the LSB (Least Significant Bit). The SRF2 computation is similar to SRF1, and average the previous success ratio with the current success ratio.
46
JournalofComputerScience,Volume1,Number1,June2007,ISSN19946244
SFR2 =
Psuccess2 / Ptotal + Psuccess1 2
Similarly, the degrading process D is calculated for a node A by subtracting the Pthreshold from the new calculated success factor ratio as D = SFRA - Pthreshold If the If D 0 then, the trust factor of the node is incremented by 1. If SRF2 exceeds Pthreshold but does not exceed by predefined value then trust factor is not incremented. Hence, notation for calculating the trust factor is T [ A, B, T, F ] = X For secured routing, we want to ensure that a malicious node can not be trusted by evaluating once. If its consecutive success ratio is satisfactory only then it is trusted as a secured neighbor [detection is more essential.] Hence, the upgrading is slower than degrading. A node, which is not malicious may have suffered by this technique. The above notation implies that the trust factor of a node A at time t with previous success ratio f is x measured by node B. V. CONCLUSION In this paper, we proposed a trust factor upgrading technique of nodes in MANET. Here, both pre-existing knowledge and direct interaction among nodes in the network can be taken into account. In our scheme, upgrading technique may not allow a secured neighbor at the first time when route is discovered as its previous trust level will be zero. But it ensures that no malicious node entering in a network will be trusted. As the next step of developing a general trust model, we plan to extend our trust model to a general form to overcome this problem and we will also look further at some issues like avoiding count to infinity problem. VI. REFERENCE [01] [02] [03] [04] [05] [06] [07] [08] [09] [10] [11] C. Perkins and E. Belding-Royer and S. Das, 2002. Ad hoc On-Demand Distance Vector (AODV) Routing , RFC 3561, July 2003. Qifeng Lu, Dec 15, 2002. Vulnerability of Wireless Routing Protocols. Jugal Krishna Das, Shareeful Islam and Abu Raihan Mostofa Kamal, An analysis and solution of security vulnerability of AODV routing protocol in Mobile Ad Hoc Networks. Lyes Guemari, August 20, 2001. An OPNET model implementation for Ad-hoc On demand Distance Vector Routing Protocol. Giovanni Vigna, Sumit Gwalani, Kavitha Srinivasan, Elizabeth M. Belding-Royer, Richard A. Kemmerer, 2004. An Intrusion Detection Tool for AODV-based Ad hoc Wireless Networks. Tony Larson, Nicklas Hedman, 1998. Routing Protocols in Wireless Ad-hoc Networks- A Simulation Study. Ian D. Chakeres, Elizabeth M. Belding-Royer. AODV Routing Protocol Implementation Design . Nikola Milanovic, Miroslaw Malek, Anthony Davidson, Veljko Milutinovic. Routing and Security in Mobile Ad Hoc Networks. Karl Levitt. Intrusion Detection for Mobile Ad Hoc NetworksInternet. Charles E perkins, 2001. Ad hoc Networking. Peng Ning, Kun Sun, 2003. "How to Misuse AODV: A Case Study of Insider Attacks against Mobile Adhoc Routing Protocols," in Proceedings of the 4th Annual IEEE Information Assurance Workshop, pages 60-67, West Point, June 2003. Stephen Carter, Alec Yasinsac. Source Position Aided Ad Hoc Routing. Kwan-Wu Chin, John Judge, Aidan Williams and Roger Kermode. Implementation Experience with MANET Routing Protocols. L. Zhou, and Z. Haas, Securing ad hoc networks, IEEE Network Magazine, vol. 13, no.6, Nov/Dec 1999. Mohamed M. Abou El Saoud. MANET reference configurations and evaluation of service location protocol for MANET. Yan, Z., P. Zhang, and T. Virtanen. Trust Evaluation Based Security Solution in Ad Hoc Networks in Proceedings of the Seventh Nordic Workshop on Secure IT Systems 2003. 2003. Norway. Virendra, M., et al. Quantifying Trust in Mobile Ad-Hoc Networks in International Con-ference on Integration of Knowledge Intensive Multi-Agent Systems, 2005 (KIMAS '05). 2005. Waltham, Massachusetts, USA: IEEE.
[12] [13] [14] [15] [16] [17]
47
Khondaker Abdullah-Al-Mamun et al

Khondaker Abdullah-Al-Mamun1, Abu Hasan Md. Sayeed2, Sabina Yeasmin3 and Farhana Sarker4
ABSTRACT
A Novel Segmented Display For Arabic Numerals. Among all characters of different languages, it is undoubted that the numerals are the most extensively used characters in real life. Numerals need to every human in their whole life. It becomes unavoidable to do any tasks or exchanges without the help of numerals. In the universe, decimal code (zero to nine) is the most widely used numerals to represent numeric values. Most of the digital equipments have some means to display the different information in a form so that one can easily understand the result or output. Usually this information is numerical data but also can be alphanumeric (numbers and letters). One of the simplest and most popular methods for displaying English numerals is 7-Segment display model that display 0 to 9 and also sometimes display the hex characters A to F. In this paper we worked on representing Arabic numerals and also proposed a unique and novel 9-Segment display model to display Arabic numerals perfectly.
Keywords: 9-Segment Display, Arabic Numerals, Segmented Display, Combination Vector, Truth table, K-map. I. INTRODUCTION Digital devices like computers, calculators etc. need to use numerals to display the numeric values. Different devices represent characters using Dot Matrix or Segmented display system. The 7-Segment display is a very common output device that uses the English decimal numbers (0-9). There are two types of display technology that are used to display numerals: LED and LCD. Though LCD consumes less power but LED looks brighter and consistent in the dark. As a result we consider our segments as LED segments. The proposed model uses almost equal Segments so the display seems uniform. Arabic Numerals have more curved corners with respect to their counterpart English Numerals. The display performance of different models mainly depend on the perfection of the curved corners appearances. It is well known that the number of segments needed to represent numerals increase linearly with the required degree of accuracy of the curved edges [1]. Mohammad Osiur Rahman et al. [5] in 2003 proposed a segmented display model for Arabic numerals but the problem is that they do not use uniform sized segments. As a result, the brightness of the model suffered from uniform intensity. Besides, some other works [2-4] also suffered from unequal size of segments. On the other hand the proposed model uses almost uniform segments and looks simple to understand and recognize. The distinguished features of the proposed algorithm are: it uses almost uniform sized segments to display all the numerals; the model uses only 9 segments; the model does not have any intersections among the segments; its architecture is very simple; and finally this model will be easier to implement. II. PROPOSED 9-SEGMENT DISPLAY After analyzing all Arabic numerals we got the architecture of our proposed segmented display. The proposed uniform 9Segment display model is depicted in Figure 1. We have tried a lot to use bar type and uniform size segments as much as possible. Matrix points those make up an oblique shape are considered as the points to be connected together to make up a segment. The segments are non-overlapping and labeled with specific alphabet.
48
Offline Divisible E-Cash
Figure 1: Proposed 9-Segment Display Model
The proposed model consists of nine segments entitled as a, b, c, d, e, f, g, h, i. The model looks easier and presentable. The major advantage of this model is non-overlapping segments and uniform segments. Besides it uses only 9 segments.
49
Md. Musfique Anwar et al.
III. REPRESENTATION OF ARABIC NUMERALS USING PROPOSED MODEL In Table 1 shows the Segmented-Layout of the proposed model. Now we see the layout of Arabic Numerals formed by the proposed model. Figure 2 shows the layout/pattern of the Arabic numerals designed by the proposed 9-Segment display model. It looks visible, accurate, and easily identifiable. This is the novel characteristics of the proposed model.
Figure 2: Segments layout for Arabic numerals
In the following, the detail analysis of the model, Truth Tables, Block Diagrams etc are furnished. Table 1 furnishes the English numerals, Arabic numerals, and also proposed models Arabic numerals. Also shows the numerals in such a way that anyone can easily compare between exact Arabic numerals with proposed models Arabic numerals and can identify closely. In spite of its huge curvature corners, our proposed model can perform better.
Table 1: Representation of Arabic numerals
Digit 0 1
Arabic
Pattern
Digit 5 6
Arabic
Pattern
IV. DESIGNING THE REQUIRED HARDWARE Since the proposed model has already been described now it is needed to determine the hardware design that will be needed to implement it. This section will also express how easily this model can be derived. First we determine the combination vector of activated segments in Table 2. That means determination of segment list required to display the different Arabic numerals. It shows the active segments list for each numeral.
Table 2: Combination Vectors for Arabic numerals Digit value 0 1 2 Arabic numeral Combination Vector for Arabic numeral { a, b, g } {i} { c, d } JournalofComputerScience,Volume1,Number1,June2007,ISSN:19946244
. 50
3 4 5 6 7 8 9
{ c, d, h } { e, g } { a, c, h } { h, i } { b, g } {a} { f, h, i }
After determining active segments it requires to determine the Truth Table for the proposed model. Table 3 shows the truth Table of the proposed model. Since computer uses BCD presentation to deal with numerals, so the BCD expression is also enclosed here. Afterwards it will be needed to determine the Sum of Products (SOP) for each Segment. Table 4 shows the SOP for each segment.
Table 3: Truth Table for the Arabic numerals Arabic Digits BCD input A 0 0 0 0 0 0 0 0 1 1 B 0 0 0 0 1 1 1 1 0 0 C D a 0 0 1 1 0 0 1 1 0 0 0 1 0 1 0 1 0 1 0 1 1 0 0 0 0 1 0 0 1 0 b 1 0 0 0 0 0 0 1 0 0 Segments c d e 0 0 1 1 0 1 0 0 0 0 0 0 1 1 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 f 0 0 0 0 0 0 0 0 0 1 g h 1 0 0 0 1 0 0 1 0 0 0 0 0 1 0 1 1 0 0 1 i 0 1 0 0 0 0 1 0 0 1
0 1 2 3 4 5 6 7 8 9
Table 4: Sum of Product for each Segment
Segment Sum of Product Function ( 0, 5, 8 ) a ( 0, 7 ) b ( 2, 3, 5 ) c d ( 2, 3 ) (4) e (9) f ( 0, 4, 7 ) g ( 3, 5, 6, 9 ) h ( 1, 6, 9 ) i Deriving SOPs it will be required to minimize the Boolean expression for each segment using Karnaugh-map (K-map) method. The K-map is a graphical tool used to simplify a logic equation or to convert a Truth Table to its corresponding logic circuit in a simple or dearly process. The expressions are shown in Table 5.
Table 5: Minimized Boolean Expression for each Segment
Segment a b c d e f g h i
Minimized Expression BCD+BCD BCD+ABCD BC+BCD BC BCD AD ACD+BCD AD+BCD+BCD+BCD BCD+BCD
Now the required detail circuit diagram and block diagram are derived for the proposed 9Segment model. Figure 3 and 4 has shown the circuit and block diagram of the proposed model respectively.
51
Md. Musfique Anwar et al.
Figure 3: Decoding Logic Circuits for different Segments
Figure 4: Block Diagram of Proposed Model
V. SPECIAL FEATURES OF THIS MODEL The model described in this paper is one of the easiest and simplest forms to represent the Arabic numerals. This model applied an empirical approach to find out the required numbers of segments and also the shapes of the individual segments. Actually it uses dot matrix approach to determine the architecture of this model. Ultimately this model uses almost uniform sized of LEDs. As a result the intensity of the brightness is equal in each part of the display area. Besides, using LED is better with respect to LCD device since it is easier to implement. Another significant feature is that it does not have any intersection among the segments. The attained shapes are very close to the original Arabic numerals. It becomes very successful to deal with the curvature part of the numerals. The required circuit diagram is also very simple and easy. Considering all these issues it can be concluded that the proposed model is an excellent model for representing Arabic numerals. The direct fair comparison is not possible due to the lack of similar experiments. As per our best knowledge this model is optimistic, easy to implement, and better in all respects. VI. CONCLUSION This paper proposed an excellent, unique, and novel 9-Segment display model to display Arabic numerals. Irrespective of being a segmented display model, the overall outlook of this model becomes lucrative, technically sound and easily understandable. As a result in future it is possible to interface/connect this model with computers or any other devices. Further development can be possible to use this approach in image processing. Whenever a numeral input come into a computer as image, first convert that numeral into dot matrix form and thereafter minimizing this form and make this into a few number of segments (according to our proposed model). Thus, the minimal numbers segmented presentation will be very easy to recognize. Besides, the memory requirement for matching algorithm of Arabic numerals will be reduced and required time reduces into very short. So in future we will try to implement this model for characters in different languages. . 52
VII. REFERENCES [1] Ahmed Yousuf Saber, Mamun Al Myrshed Chowdhary, Suman Ahmmed and Chowdhury Mofizur Rahman: Designing 11-segment display for Bangla Digits, Proc. 5th Int. Conference on Computer and Information Technology (ICCIT), Dhaka, Bangladesh, (2002) p.237. [2] S. M. Masum, M. S. Islam Khan, SK. Md. Mizanur Rahman and M. Ali: Designing 10-Segment Display for Bangla Digits, Proc. 3rd Int. Conference on Electrical Electronics and Computer Engineering (ICEECE), Dhaka, Bangladesh, 2003, p.333. [3] A. K. M. Khaled Ahsan Talukder and K. Roy: Development of 16-Segment Multilingual Display Driver, Proc. 3rd Int. Conference on Electrical Electronics and Computer Engineering (ICEECE), Dhaka, Bangladesh, 2003, p.342. [4] Fazle Rabbi, Mohammad Kabir Hossain and Monzur Ahmed: An 8-Segment Display for both English and Bengla Digits, Proc. 6th Int. Conference on Computer and Information Technology (ICCIT), Dhaka, Bangladesh, (2003) p.338. [5] Mohammad Osiur Rahman, Md. Anwarul Azim, Mohammad Sanaullah Chowdhury and Dr. Md. Nurul Islam: Different Segment Displays for Bangla, English and Arabic digits, Proc. 6th Int. Conference on Computer and Information Technology (ICCIT), Dhaka, Bangladesh, (2003) p. 299. [6] Doglas V. Hall: Microprocessor and interfacing Programming and Hardware, McGraw-Hill International Inc., 1995, p.267. [7] Morris Mano: Digital Logic and Computer Design, Prentice Hall Inc., 1996, p137, 191, 247.

Md. Musfique Anwar1 and Sohail Ahmad1
ABSTRACT
Present days electronic payment systems have a major problem; they cannot handle the security and the users anonymity. Ecash is a payment system, which enables a secure off-line transaction without revealing the payers identity. E-cash can be used both as paper cash and electronic money since it keeps its users anonymity, enables off-line transactions, is portable and at the same time offers the ability of electronic transactions. This paper explains the concept of e- cash and discusses its properties. It also discusses the implementation and practical usage of e- cash. When implementing the protocol, CORBA (Common Object Request Broker Architecture) is used for distributed environment. Binary tree approach is used to construct a divisible E-cash. J2ME (Java 2 Micro Edition) is used for the implantation of client part, Bouncy Castle Crypto Package for security protocol and C++ is used for the server part. Keywords: E-Cash, Divisibility, CORBA, ORB, J2ME and Bluetooth.
I. INTRODUCTION Essentially, E-cash mimics the functionality of paper cash. More technically, E-cash is a payment message bearing a digital signature which functions as a medium of exchange or store of value. Paper currency and coins represent value because they are backed by a trusted third party, the government and the banking industry. Digital coins will also represent value because they are backed by a trusted third party, usually a bank that is willing to convert E-cash to physical cash. There are a number of electronic cash protocols. To a degree, all E-cash schemes operate in the following manner: A user installs a "cyber wallet" onto computer. Money can be put in the wallet by deciding how much is needed and then sending an encrypted message to the bank asking for this amount to be deducted from the user's account. The bank reads the message with private key decryption and verifies if it has been digitally signed in order to identify the user. The bank then generates "serial numbers", encrypts the message, signs it with its digital signature and returns it. The user is now entitled to use the message (coin or token) to spend it at merchant sites. Merchants receive E-cash during a transaction and see that it has
53
Md. Mahmud-uz-zaman et al.
been authorized by a bank. They then contact the bank to make sure the coins have not been spent somewhere else, and the amount is credited to the merchant's account. In recent there are much research has been performed in the area of off-line electronic currency. These protocols proposed to the consumer to enabling to withdraw "electronic coin" from a bank and to later spend these "electronic coins" at a shop or anywhere in an "off-line" manner. Here, off-line refers to the property that communication with a bank or authorized center is unnecessary during the payment protocol. A "divisible" coin means that suppose I have some amount of money, say $x, I can divide these money in several small value and that can be spent several times as long as the sum total of all it's the transactions does not exceed $x. So far, only Okamoto and Okta have shown the construction of a divisible off-line electronic cash system in [1]. Brands touched upon the possibility of realizing divisibility with his scheme [2], but did not achieve it in the sense of [1]. Pallies divisible coin construction [3] is efficient. The communication complexity during payments is linear in N where N = (the total coin value) / (minimum divisible coin value). A system in which a coin worth $367 consists of 367 $1 coins is a rather unwieldy and inefficient divisible cash system. In contrast the communication complexity of payments in is of the order of log N. 1. Basic Framework In the following section, a basic framework for divisible E-cash is stated. First, basic notations used in the explanation of E-cash scheme are introduced. 1.1 Notation and assumptions Let p and q be large primes such that q | (p-1). The notation x R, X means that x is uniformly and randomly selected from the X. Also let || denote concatenation and let |x| be the length of the binary representation of x. If b is a bit, let b represent the negation of b. Let H:{0,1}* {0,1}2|q| be a polynomial time computable one way hash function. 1.2 Overview of Divisibility Binary tree approach is used as [1] did. Each coin of worth w is associated with a tree of (1+log w) levels and w leaves. The tree can be thought of as a collection of w paths of length (1+log w), each originating from the root and terminating in a leaf. Each node of the tree represents a certain denomination. The root node is assigned a monetary value of w, and the value of all other node's parent. With this tree, it will show that for a single coin of worth w, it will be w possible for a consumer to engage in several transactions, such that the sum total of the amounts of each transaction is less than or equal to w . II. REMOTE COMMUNICATION TOOLS 1. CORBA Common Object Request Broker Architecture (CORBA), is in industry consensus standard that defines a higher level facility for distributed computing. CORBA standardizes interfaces and semantics for object-oriented middle ware. It includes a specification for the Object Request Broker (ORB), a software library with standardized CORBA object interfaces that allows clients and targets to communicate with each other across a network in a well-defined way. In addition, CORBA automatically applies a range of useful services to communications. After the ORB is initialized, all CORBA objects can be invoked by applications like local software. 1.1 CORBA Architecture Overview CORBA is an object-oriented architecture. CORBA objects exhibit many features and traits of other object-oriented systems, including interface inheritance and polymorphisms. The Object Request Broker (ORB) CORBA is a specification for an application-level communication infrastructure. It provides communication facilities to application through two mechanisms: static interfaces and the Dynamic Invocation Interface (DII).An Interface Repository stores on-line description of known OMG[4] IDL interfaces, any interface can be used with either mechanism. An ORB is a software component whose purpose is to facilitate communication between objects. It does so by providing a number of capabilities, one of which is to locate a remote object, given an object reference. Another service provided by the ORB is the marshaling of parameters and return values to and from remote method invocations. Recall that the Object Management Architecture (OMA) includes a provision for ORB functionality; CORBA is the standard that implements this ORB capability.
54
An Efficient Algorithm for Reduction of Computational Time in Parallel Implementation of NWS Algorithm
When an application component has obtained a reference to an object whose services the component wants to use, that component can invoke methods of that object. Generally, these methods take parameters as input and return other parameters as output. Another responsibility of the ORB is to receive the input parameters from the component that is calling the method and to marshal these parameters. A product of the marshaling/unmarshaling process is that, because parameters are converted upon transmission into a platform-independent format and converted into a platform-specific format upon reception, the communication between components is platform-independent. This means that a client running on, for instance, a Macintosh system can invoke methods on a server running on a UNIX system. In addition to independence of operating system used, differences in hardware are also rendered irrelevant because the ORB automatically makes these conversions as necessary. 2. J2ME J2ME is a highly optimized Java runtime environment. J2ME is aimed at the consumer and embedded devices market. This includes devices such as cellular telephones, Personal Digital Assistants (PDAs) and other small devices.
Fig 1 High level view of J2ME Figure 1 shows the J2ME architecture. Java 2 Standard Edition (J2SE) developers should be familiar with Java Virtual Machines (JVMs) and at least one host Operating System (OS). The OS will vary on different mobile devices. Some devices run the Symbian OS, others run some other OS developed by the manufacturer. 2.1 Configurations and profiles Mobile devices come with different form, features and functionality, but often use similar processors and have similar amounts of memory. Therefore configurations were created, defining groups of products based on the available processor power and memory of each device. A configuration outlines the following[5]: i) The Java programming language features supported ii) The JVM features supported iii) The basic Java libraries and Application Programming Interfaces (APIs) supported.
There are two standard configurations for the J2ME at this time, Connected Device Configuration (CDC) and Connected Limited Device Configuration (CLDC). The CDC is targeted toward powerful devices like Internet TVs and car navigation systems. The CLDC is targeted toward less powerful devices like mobile phones and PDAs. The vast majority of Java enabled mobile devices available to consumers today use CLDC. A profile defines a set of APIs which reside on top of a configuration and offers access to device specific capabilities. The Mobile Information Device Profile (MIDP) is a profile to be used with the CLDC and provides a set of APIs for use by mobile devices. These APIs include classes for user interface, persistent storage and networking. MIDP applications are called MIDlets. 2.3 Bluetooth Technology
JournalofComputerScience,Volume1,Number1,June2007, ISSN : 1994-6244
55
Bluetooth is a low cost, low power; short-range radio technology intended to replace cable connections between cell phones, PDAs and other portable devices. The Bluetooth specification aims to allow Bluetooth devices from different manufacturers to work with one another, A piconet is the usual form of a Bluetooth network and is made up of one master and one or more slaves. When connecting two piconets the result will be a scatternet. III. E-CASH PROTOCOL 1. Construction This section outlines the various protocols in divisible electronic coin scheme. When the customer wants to withdraw $x from the account, an electronic coin of worth $x is then obtained by executing the withdrawal protocol with the bank. 1.1 Divisibility Divisibility means the ability to make change. A divisible coin worth some amount money, say $x, is a coin that can be divided in several small value and that can be spent several times as long as the sum total of all its the transactions does not exceed $x. Example: For $4 level 3
Figure: 2 Divisibility of electronic coin 2. Withdrawal of E-cash by user U0 User(Ui) Setup2() precomputation stage Calculate2() Bank(B) trans1(I0) trans2(z0, a0, b0) - - -trans3(c0) Calculate4() trans4(r0) - Calculate3() Calculate1()
Fig 3 Withdrawal of E-cash by a user u0 Setup 1() 1. 2. 3. Let p, q, g, g1, g2 be the system parameters published by the bank where the orders of g, g1, g2 is q. x and x1 are the secret key of Bank. Here, x is used for issuing E-cash and x1 is used for issuing coin extension. h = gx, ht = gxt are the public keys of Bank.
Setup 2() 1. 2. u0 be U0s secret key. I0 = g1u0 be the identity of user U0.
Precomputation stage 1.
The customer wants to withdraw wl = 2l.

56
2. 3.
The customer selects a random value e as a secret seed value. For a node: n0j1j2jv, ji {0,1} 0j1j2jv, = H(e || 0j1,j2..jv) From a nodes - value, we define its 2 values of leaf node 0j1j2jv = 0j1j2jv,1 || 0j1j2jv,2 - values and values for internal nodes 0j1j2jv0 = 0j1j2jv0,1 || 0j1j2jv0,2 0j1j2jv0 = 0j1j2jv1,1 || 0j1j2jv1,2
Then its value is defined as 0j1j2jv = H (H(g1 0j1,j2..jv0,1 g2 0j1,j2..jv0,2 mod p) || H(g1 0j1,j2..jv0,1 g2 0j1,j2..jv0,2 mod p)) U obtains, T = g1 0,1 g2 0,,2 mod p(note 0 = 0,1 || 0,2 ) Calculate 1() B subtracts wl = 2l dollars from U0s account balance. B forms m0 =I0 g2 mod p. B chooses w0 R Z*q. B calculates z0 = mx0 mod p. a0 = g0w0 mod q mod p. b0 = m0w0 mod q mod p. Calculate 2() 1. 2. 3. 4. 1. 2. 3. 4. 5. 6. 7. U0 generates s0, t0, v0 R Z*q. m'0 = g1u0s0 g2s0 mod p. z'0 = z0s0 mod p. a'0 = a0t0 gv0 mod p. b'0 = b0s0 t0 (m'0) v0 mod p. c'0 = H(m'0 , z'0 , a'0 , b'0 , T ). c0,= c'0 / t0,
Calculate 3() 1. r0 = c0x + w0 mod q.
Calculate 4() Calculates r0, = r0 t0 + v0 after verifying that m0r0 = z0 c0b0 mod p gr0 = h c0a0 mod p The user obtains the signature Sign (m'0, T) = { z'0, a'0, b'0, r'0} Verifies the signature gr'0 h c'0a'0 (mod p) (m'0) r'0 (z'0) c'0b'0 (mod p) c'0 H (m'0,T,z'0, a'0, b'0) IV. RESULT AND DISCUSSION 1 Tools, Device Tools and devices that are needed are: J2ME wireless toolkit, Java Development Kit 1.4.2, Open source encryption library: The Bouncy Castle Crypto Package, Open source Java class file obfuscator: ProGuard, Bluetooth Device, Mobile Phone Sets.
57
1.1 The Bouncy Castle Crypto Package The Bouncy Castle Crypto package is a Java implementation of cryptographic algorithms; it was developed by the Legion of the Bouncy Castle. Bouncy Castle is an open source encryption library; there is a lightweight version suitable for use with J2ME. This package is needed to use BigInteger class of java.math package. 1.2 ProGuard ProGuard is a Java class file shrinker and obfuscator. It detects and removes unused classes, fields, methods, and attributes, and it renames the remaining classes, fields, and methods using short meaningless names. The resulting jars are smaller and harder to reverse-engineer. ProGuard can also be used to list unused fields and methods in an application, and to print out the internal structure of class files. 1.3 Java APIs for Bluetooth Wireless Technology Recent developments in the Java programming language make it possible for Java developers to write applications for wireless devices such as cell phones and personal digital assistants. In 2000, the Java programming community [6] recognized the importance of creating a standard extension to the Java programming language for use with Bluetooth devices. A standard application programming interface (API) for Bluetooth was needed because each Bluetooth software protocol stack had its own API for application programmers. These proprietary APIs meant that a Bluetooth application had to be ported to different Bluetooth stacks to run on different devices.
Fig 4 Sample output V. FUTURE EXTENSION AND CONCLUSION 1 Future Extension In future, an observer will be added with all clients. The central problem of off-line coins is the double spending i.e. missing of originality of the coin. If a client spends the same coin more than one time then it is known as double spending. An observer can prevent double spending in starting. That means no client can double spend any coin because the observer will check it. The use of observer is a kind of first line defense. If the client cannot manipulate the device the observer can prevent double spending. If the succeeds in tampering the observer, the double spending detection identifies the user afterwards. The client have to break the observer physically and the electronic cash system mathematically to cheat the bank. VI. CONCLUSION The main purpose of this paper is to implement E-cash in a distributed environment. CORBA is used for distributed environment and Bouncy Castle Crypto Package for the implementation of big integer which supports a huge range of integer variable such as secret key and public key of bank for security. ProGuard is used which is a Java class file shrinker and obfuscator & it detects and removes unused classes, fields, methods, and attributes, and it renames the remaining classes, fields, and methods using short meaningless names. As the size of the jar and jad file is smaller it wills little memory when it will be inserted into a mobile. In future, I will try how to use CORBA in J2ME for distributed environment and try to port the
58
client part code into a Linux mobile phone. In that case, a client can easily communicate through his Linux mobile phone to a server. VII. REFERENCES T. Okamoto and K. Ohta, Universal electronic cash, in Advances in Cryptology-Crypto 91, pp. 321-337, 1991. S.Brands, Untraceable off-line cash in wallet with observer, Proceedings of CRYPTO, pp. 303-318, 1993. J. Pailles, New protocol for electronic money, Proceedings of the Auscrypt, pp. 263-274, 1992. SAMs publishing, Teach Yourself CORBA In 14 Days. http://docs.online.bg/DB/Teach_Yourself_CORBA_In_14_Days/copy.htm. [5] J2ME: http:// java.sun.com/j2me [6] C Bala Kumar, Paul J. Kline, Timothy J. Thompson, Bluetooth Application Programming with the JAVA APIs, Elsevier, 2004.
[1] [2] [3] [4]
Md. Mahmud-uz-zaman1, Md. Shoyaib2 and Md. Mahbubul Alam Joarder3
ABSTRACT
Needleman Wunsch Seller (NWS) Dynamic programming algorithm can be implemented in parallel, which reduces time complexity. But in this parallel implementation number of processors needed increases with the sequence lengths. Thus ideal parallel implementation is practically impossible. But if some cells for computation are omitted choosing the best paths to traverse, number of processor needed can be reduced. In this paper an algorithm is developed to calculate the best path by only utilizing limited resources. Keywords: Parallel Processing, Time Reduction, Dynamic Programming.
I. INTRODUCTION Computer processor speed increased dramatically from early 8088 processor to Pentium, so it was thought that increasing processing speed could solve biological sequence analysis and searching problem. But contemporary genome projects have delivered a blow to this early confidence. From the completion of the first whole organism's genome [1], the growth rates for biological data have become a determinant to sequential computing processing capability. At this point, sequential (one-processor) computing can allow only a small part of the massive, multidimensional biological information to be processed. Under this scenario, comprehension of the data and understanding of the data-described biological processes could remain incomplete, causing us to lose vast quantities of valuable information because CPU-power and time constraints could fail to follow critical events and trends. From a computational point of view, there are several ways to address the lack of hard computing power for bioinformatics. The first is by developing new, faster heuristic algorithms that reduce computational space for the most timeconsuming tasks [2]. The second is incorporating these algorithms into the ROM of a specialized chip [3]. The third and most promising consideration, however, is parallel computing. Two or more microprocessors can be used simultaneously, in parallel processing, to divide and conquer tasks that would overwhelm a single, sequential processor. However, promising parallel computing still requires new paradigms in order to harness the additional processing power for bioinformatics. In parallel processing there can be limitation of resources as its need increases with sequences lengths. The goal of this paper is to reduce the resource needed to maintain the accuracy. The rest of the paper has been arranged into several sections. Section 2 describes basic concept of NWS algorithm. Section 3 proposes an algorithm for time complexity reduction. Section 4 shows results of this algorithm. Conclusion is given in section 5.
59
Health Hazards Of Cellular Phones: Bangladesh Perspective.
II. BACKGROUND For global alignment Needleman Wunsch algorithm [4] is widely used. Needleman Wunsch Seller (NWS) does the same global alignment with a different scoring pattern. NWS calculates an alignment which minimizes the difference score between two sequences. We can assign a difference score d to each alignment,
d = s + gp ( gs i )
i =1
where s is the number of substitutions, n is the number of gaps,
gsi is the size of gap i, and gp is a gap penalty function
assigning positive values to all gap sizes. It assumes that gap penalties grow linearly with gap sizes. Thus,
gp ( gs ) = gs gap _ penalty
where gap _ penalty > 1 . An optimal alignment with respect to a difference score is an alignment, which possesses the lowest difference score. The NWS algorithm calculates an optimal alignment between alignments between all prefixes of their difference scores.
S1 and S 2 by memorizing optimal
S1 and S 2 . Two matrices may be allocated for storing alignments between prefixes and S1 and the first b elements of S 2 ,
score _ matrix[ a, b]
contains the difference score of an optimal alignment between the first a elements of
path _ matrix[ a, b] contains the index of the previous cell in the alignment. The alignment is constructed by following the indices stored in path-matrix. The 0th row corresponds to alignments possessing gaps at the beginning of S1 . Thus, score _ matrix[0, a ] = gp ( a )
Similarly, the 0th column corresponds to alignments possessing gaps at the beginning of Ss. Therefore,
score _ matrix[b,0] = gp (b)
The remainder of the matrix is calculated inductively using the formula: score _ matrix[i, j ] = min (
score _ matrix[i 1, j 1] + ) dis tan ce(S 1 [i ], S 2 [ j ])
score _ matrix i 1, j + gap _ penalty , score _ matrix[i, j 1] + gap _ penalty ,

(1)
where distance ( x, y ) = 0
=1
if x = y if x y
The indices of the previous cell are stored in path _ matrix[i, j ] . After the matrices have been completely determined, an optimal alignment between the entire sequences is constructed by following indices starting at path_matrix[ S1 , Time and space complexity is equal to the multiplication of we tried to improve the time requirements by reducing the number of cells be computed leading to best scores. III. PROPOSAL OF AN EFFICIENT ALGORITHM The work is to calculate the matrix final value or score with fewer processors. But The main challenge is to get the score at any position. As we know in dynamic programming any cell score S[ i, j] is dependent on previous upper and left cells.
S1 and S 2 that is ( S1 S 2 ) . In this paper work
S 2 ].
Figure 1: A cell value is dependent on all the upper and left cell values S[ i any So finding a cell value at, j] position randomly will depend on the upper and left cell values. NWS algorithm minimizes the difference between the alignments. So, at any position in a matrix the alignment score comes from upper, left or corner . But if calculate through wavefronts [5] we have to consider only the previous two wavefronts as values are coming from previous two wavefronts or upper three cells. And the optimal score or optimal alignment should come from the optimal
Mohammad Naderuzzaman et al.
alignment up to that point. If could go in forward direction and if that direction in an alignment path is right then we actually dont need to calculate other cells values those are not important. Here we wanted to move from [0,0] to [l1 ,l 2 ] in the matrix through the wavefronts. The final version can be processed in three stages losing some accuracy. First proposal First step of this algorithm is to divide the total number of wavefronts into some wavelevels. Wavelevel is a user defined positive number, which is taken as a stoppage. At a wavelevel, the minimum score is searched on that and previous wavefront (two wavefronts are taken into account as any alignment can also come only from the last two wavefronts). let S1 and S 2 are two sequences of length n1 and n2 . Total Wavefronts w need to compute are n1 + n2 1 . Wavelevel l is a finite number ( l < w ). And total computation will be done in w / l steps or in w / l wavelevels Node computations will go to first wavelevel l and then find the min score. Next it will again go for computation from that point to next wavelevel.
w =2 Step 1 w =l Figure 2: Step wise computation in wavefronts Step =2 , w=2*l For Single processor node computation will be (w / l )(l + l 1 + ... + 1) .For multi processor Time complexity would be O( n1 + n2 1 ) and maximum processor needed would be l. Next step improvement If at the wavelevel from the last two adjacent wavefronts, there are more that one minimum scores, then we would miss one best alignment that could go from that best score. Here we tired to incorporate all the best scores that can become a best alignment.
i
Figure 3: considering best points At any wavelevel there can be more than one best point. We can summarize these points to two best points, one is upper and one is lower. In upper best point j value will be highest and i value will be lowest and on contrary in lower best point i value will be the highest and j value will be the lowest. Let the best lower points i = lBi and j = lB j , higher best points i = hBi and j = hB j . At next wavelevel, computational nodes will be from lBi + l , lB j to hBi , hB j + l . Comparing with previous one, the main variation is diff which the difference between high and low best points. For Single processor node computation will be ( w / l )((l + diff ) + ..diff ) . For multi processor Time complexity would be O( n1 + n2 1 ) and maximum processor needed would be l+diff. While we have considered all the best scores in our computation, still there can be efficiency loss. There can be subtle scores that is very closer to the minimum score and that can become the best alignment. That score alignment passes from the left corner or right corner pockets.
Left Corne Figure 4: Calculation of corner area values in wavefronts

If consider two corner pockets, then all the scores that can contribute to the best alignment are coming into attention. And new wave level is still from the best points. From the previous one the main difference is left and right corner pockets are taken into account. In result, all the alignments that could pass the left or right pockets are now taken in consideration. Thus the chance of getting the best alignment from the best score has begun to be greater than the previous views. In figure 5 an algorithm is drawn. Line 1 and 2 are the initialization part. wave _ level is a user defined value which is an experimental value also. Total wavefront number is from line 2[4]. Lines 3 to 26 are the main computation parts. Where w varies upto wavefront _ max . in line 4 1 getting the wave_level and initialize variables and matrix 2 wavefront_max=len_a+len_b-1; 3 for( w= 1 to wavefront_max){ 4 waves++; 5 lilowval=preStepLowI+((waves-1)/2);//left side of Original 6 ljlowval=preStepLowJ+((waves)/2); 7 if(lilowval>len_b){ 8 ljlowval=(lilowval-len_b)+preStepLowJ; 9 lilowval=len_b;} 10 rimaxval=preStepHighI+((waves)/2);//Right side of Original 11 rjmaxval=preStepHighJ+((waves-1)/2); 12 if(rjmaxval>len_a){ 13 rimaxval=(rjmaxval-len_a)+preStepHighI; 14 rjmaxval=len_a;} 15 ilowval=lowI+(waves-1); 16 imaxval=highI; 17 jlowval=lowJ; 18 jmaxval=highJ+(waves-1); 19 if(lilowval>ilowval && ljlowval<jlowval){ 20 for (i = lilowval; i >= ilowval; ) for(j = ljlowval; j <= jlowval; j++,i--){
D[i - 1][j - 1] + similarity (seq_a[j - 1], seq_b[i - 1]) D[i ][ j ] min D[i - 1][j] + gap_val D[i][j - 1] + gap_val
21 if(rimaxval<imaxval && rjmaxval>jmaxval) { printf("Some works in rightside\n"); for (i = imaxval; i >= rimaxval; ) for(j = jmaxval; j <= rjmaxval; j++,i--)
D[i-1][j-1]+ (seq_a[j1],seq_b[i 1]) similarity D[i][ j] minD[i-1][j]+ gap_val D[i][j-1]+ gap_val

22 for (i = ilowval; i >= imaxval; ) //in the middle portion for(j = jlowval; j <= jmaxval; j++,i--){
D[i - 1][j - 1] + similarity (seq_a[j - 1], seq_b[i - 1]) D[i ][ j ] min D[i - 1][j] + gap_val D[i][j - 1] + gap_val
23 if(waves=wave_level or waves=wave_level-1){ 24 get_minScores and set preStepHighI , preStepHighJ preStepLowI and preStepHighI 25 if(waves==wave_level){
set highWaveI, highWaveJ, lowWaveI, lowWaveJ waves=1;} 26 } Figure 5: Algorithm for final view is increased. waves varies from 1 to wave _ level . In lines 5 to 18 4 points are determined. These are ( lilowval , jlowval ), ( ilowval , jlowval ), ( imaxval , jmaxval ) and ( rimaxval , rjmaxval ). To compute these values ( preStepLowI , preStepLowJ ) and ( preStepHighI , preStepHighJ ) points values are very important. These are the
waves
previous wavelevels lower point and upper points. ( lilowval , jlowval ) point will be derived from ( preStepLowI , preStepLowJ )and ( rimaxval ,s rjmaxval ) will be derived from ( preStepHighI , preStepHighJ ). Another two points ( lowI , lowJ ) and ( highI , highJ ) are important for deriving ( ilowval , jlowval ) and ( imaxval , jmaxval ). ( lowI , lowJ ) and ( highI , highJ ) are the lower and higher points of the minimum scores in the last two wavefronts at any wavelevel. There might be a cutting situation. At lines 7 and 12 two cutting condition is taken into account. If ( lilowval , jlowval ) or ( rimaxval , rjmaxval ) go greater than len _ a or len _ b , then it is adjusted in the matrix. In 18 to 22 the computational work is being done. 19,20 is for left side computation, 21 for right side and 22 for center computations. left side nodes are from ( lilowval , jlowval ) to ( ilowval , jlowval ). right side computation will be from ( imaxval , jmaxval ) to ( rimaxval , rjmaxval ) and center computation will be from ( ilowval , jlowval ) to ( imaxval , jmaxval ). In line 23 and 24 tasks at the wavelevel is done which include finding ( preStepLowI , preStepLowJ ), ( preStepHighI , preStepHighJ ), ( lowI , lowJ ) and ( highI , highJ ). Line 25 concludes and initialize waves which means a wavelevel is completed and task of next wavelevel will start. Comparing with previous ones, node computation will on an average will be on an average will be w l which for Single processor node computation will be w 2 and for multi processor Time complexity would be same O( n1 + n2 1 ) and maximum processor needed would also be same l+diff. IV. RESULT A comparative study of time is done with the full l1 l 2 calculation version of NWS and the final view. Here two same length sequences are used. Optimal score alignment time is only considered at our final proposal code and corresponding wavelevel is noted down. The results were found in Pentium 2 350MHz single processor. Here the time shown is only for generating the distance matrix. Seq. lengths 1000 2000 3000 4000 5000 6000 Time (third view) 0.46 0.93 1.39 1.87 2.41 2.93 Time (NWS) 0.8 3.12 7.10 12.85 20.01 33.52 Wavefront
19 20 34 50 52 60 Table 1: comparative time for different sequence lengths
40 35 30 25 20 15 10 5 0
00 00 00 00 10 30 40 50 20 60 00 00
Time in sec
Time (third view) Time(NWS)
Sequence length
Figure 6: A time comparison among NWS and third view In the above figure, time needed in full matrix computation grew exponentially with the increase in length of sequences while time needed reduced dramatically and showed a steady growth in time with sequence lengths. V. CONCLUSION AND FUTURE WORKS This algorithm can be said as a heuristic algorithm, which can give approximately the best alignment at a negligible time comparing with full matrix calculation. One important aspect is guessing the wavelevel. If increase the value of wavelevel, accuracy of alignment score increases and after a certain point alignment score remains stable very close or as same as the full matrix calculated original score. One characteristics of wavelevel steps is that it is fixed for the whole wavefronts where the number of steps is w / l . Another work can be with the variable wavelevels in any step of computation, which might increase the alignment accuracy. Key identification of how many wavefronts will be taken into computation can come from the score ranges means if the scores at a wavelevel are very similar or variations are not neat then might increase the wavefronts for the next step in order to guide the best alignment path for more number of wavefronts. If the variations are large then can be clear identification that the possibility of other paths going through those higher score becomes lesser as NWS minimizes the difference between alignments. VI. REFERENCES [1] O Trelles On the Parallelization of Bio-informatics Applications. in Briefing in Bioinformatics vol. 2, no. 2, May 2001, pp. 181-194. [2] Altschul, S.F., Madden T.L., Schaffer A.A., Zhang J., Zhang Z., Miller W. and Lipman D.J., (1997) [3] AIDAN BUDD AND DAVID JUDGE, MAKING THE MOST OF YOUR MULTIPLE SEQUENCE ALIGNMENTS (MSAS), MANUAL REFINEMENT, TOOLS, CONCEPTS, APPLICATIONS, MARCH 2007, JOINTLY ORGANISED BY: DEPATMENT OF GENETICS, CAMBRIDGE UNIVERSITY, EMBL (EUROPEAN MOLECULAR BIOLOGY LABORATORY), BIOACCELERATOR AT EMBL HTTP://ETA.EMBL-HEIDELBERG.DE:8000/ [4] Needleman, S. B., and C. D. Wunsch. A General Method Applicable to the search for Similarities in the Amino Acid sequences of Two Proteins. in Journal of Molecular Biology 48. 1970, pp. 443-453. [5] Iyengar, A. K. Parallel DNA Sequence Analysis. Masters Thesis, Technical Report TR-428, Laboratory for Computer Science, Massachusetts Institute of Technology. Cambridge, MA 02139, 1988.
Health Hazards Of Cellular Phones: Bangladesh Perspective
Mohammad Naderuzzaman1, Tahsina Hoque2 and Sohail Ahmad2
ABSTRACT
The biological effects associated with exposure to electromagnetic fields used in cellular phones and mobile communication devices has grown considerably. During the last decade, many concerned organizations as well as individual scientists have taken this issue into consideration. The only media presently used for cellular communications is the electromagnetic spectrum; the result of constant exposure to this spectrum is the concern of this paper. The topics of this paper will review recent scientific evidences related to possible health hazards of mobile phone and wireless communication devices, recent regulatory developments related to human exposure to radio frequency energy and the initiatives taken by the Bangladesh Telecom Regulatory Board (BTRC). It is also an attempt to make the reader aware of threat to human life and ecosystem, caused by cellular phones and wireless devices. Keywords: health hazard, cellular Phone, Bangladesh and BTRC.
I. INTRODUCTION Computing and communications science has been drastically changed in the past decade or two. The wireless communication has been developed much as well as cables and twisted pair wires to that of wireless transmission. Many consumer, industrial products and applications make use of some form of electromagnetic (EM) spectrum. The electromagnetic energy that is of increasing importance worldwide is radio frequency (RF) energy, including radio waves and microwaves, which is used for providing telecommunications, broadcast and other services. In the United States the Federal Communication Commission (FCC) authorizes or licenses most RF telecommunications services, facilities and devices used by the public, industry and state and local governmental organizations. Heightened awareness of the expanding use of RF technology has led some people to speculate that electromagnetic pollution is causing significant risks to human health. In Bangladesh, the Bangladesh Telecommunication Regulatory Commission (BTRC) is responsible for these types of services. But so far they havent take any controlling measure for RF emission. Electromagnetic waves travel through space at the speed of light,the wavelength and frequency of an electromagnetic wave are inversely related by a simple mathematical formula: F = C or = C F Since the speed of light in a given medium or vacuum does not change, high frequency electromagnetic waves have short wavelengths and low-frequency waves have long wavelength. The electromagnetic spectrum shown in Figure- 1 includes all the forms of electromagnetic energy from extremely low frequency (ELF) energy, with very long wavelengths, to X-rays and gamma rays, which have very high frequencies and correspondingly short wavelengths. In between these extremes are radio waves, microwaves, infrared radiation, visible light, and ultraviolet radiation, in that order. The RF part of the electromagnetic spectrum is generally defined as that part of the spectrum where electromagnetic waves have frequencies in the range of about 3 kHz to 300 GHz. The frequencies ranges and its corresponding energy II. ELECTROMAGNETIC RADIATION Energy associated with the electromagnetic spectrum is given by the formula Where h is the Planks constant in electron volt (4.135667e-15 eV), f is the frequency in hertz Electromagnetic radiation can be classified as Ionizing and non-ionizing radiation. 1. Ionizing Radiation Ionization is a process by which electrons are stripped from atoms and molecules. This process can produce molecular changes that can lead to damage in biological tissue, including effects on DNA, the genetic material. This process requires interaction with quantum event. X-ray or gamma rays are examples of ionizing radiation. Ionizing radiation is also associated with the generation of nuclear energy. 2. Non-Ionizing Radiation As explained earlier, electromagnetic radiation is defined as the propagation of energy through space in the form of waves or particles. The energy associated with a photon, the elemental unit of an electromagnetic wave, depends on its frequency (or
E = hf
Md. Aman Mahbub et al.
wavelength). The higher the frequency of an electromagnetic wave, the greater will be the energy of a photon associated with it. The energy content of a photon is often expressed in terms of the unit electron-volt or eV Photons associated with X-rays and gamma rays (which have very high electromagnetic frequencies) have relatively large energy content. The photon energies associated with the various frequencies of the electromagnetic spectrum are shown in the lower scale of Figure-1.The RF energy is characterized as non-ionizing radiation, along with visible light, infrared radiation and other low bands of the spectrum. 3. Radio Frequency Measurement Because an RF electromagnetic field has both an electric and a magnetic component (electric field and magnetic field), it is often convenient to express the intensity of the RF field in terms of units specific for each component. The unit volts per meter (V/m) is often used to measure the strength (field strength) of the electric field, and the unit amperes per meter (A/m) is often used to express the strength of the magnetic field. Power density is defined as power per unit area. For example power density can be expressed in terms of miliwatt per square centimeter (mW/cm2) or microwatts per square centimeter (W/cm2). With respect to frequencies in the microwave range and higher, power density is usually used to express intensity since exposures that might occur would likely be in the farfield. III. THE HAZARDS An established effect from wireless radiation, even small, could have a considerable impact in terms of public health [1]. Even a small effect could create an epidemic size problem. Apart from wireless device that users carry with them, wireless transmission towers for radio, TV, telecommunications, radar, and many other applications too emit radio frequency radiation. The effects of radio frequency radiation do accumulate over time [2]. Another concern is that due to march toward pico-cells, greater number of base stations would be co-located and would undoubtedly result in increased level of public exposure to microwave energy over time. The permissible Exposure limits of FCC and IEEE are shown in the Table 1. f = frequency in MHz Con= Controlled Un = Uncontrolled Table: 1: FCC Limits for Maximum Permissible Exposure (MPE) IEEE (C95.1-1991) recommended MPE for cellular bands can be derived from mW/cm where f is frequency in MHz. It shows that 2.84 to 2.98 mW/Cm 6.63 mW/Cm
2 2 2
the formula
MPE = f/300
in 850/900 MHz Band and 6.43 to
in 1900MHz band are allowed exposure limit.
1. RESEARCH ON EMF EXPOSURE TO HEALTH Some laboratory or epidemiological evidence that is associated with public exposure (controlled exposure) to microwaves from cellular phone and base station antennas are related to cancer [3]. In 1995, researchers at University of Washington, Seattle, found DNA breaks in cell exposed to wireless phone radiation. Subsequent attempts by researchers at Washington University in St. Louis to duplicate the work were unsuccessful. Another epidemiological study found that right-handed people who used cell phones and had brain tumors tended to have them on the right side of the head-a result that could show a link to radiation from the phones. However, no such correlation appeared in left-handed cancer patients. It was also reported that non-ionizing radiation may speed up the cancer though it would not cause cancer. In 1998, work conducted on slice of rat brain taken from hippocampus (a structure with a role in learning), at the Defense Evaluation and Research Agencys labs (UK) showed that mobile phones could scramble memories. But in humans, the hippocampus is buried too deep in the brain to be influenced by emissions from mobile phones, say some scientists. Some experimental results conducted, show biological effects that occurred in studies of cell cultures and animals after exposures to low-intensity RFR (RF radiation) [2]. The findings are as follows: An increase in a molecular stress response in cells after exposure to a RFR at a SAR of 0.001 W/kg. This stress response is a basic biological process present in almost all animals-including humans.

64
Robertson-Walker Metric and Cosmology
An increase in calcium efflux in cells after exposure to RFR at 0.005 W/kg Calcium is an important component of normal cellular functions. An increase in the permeability of the blood-brain barrier in mice exposed to RFR of 0.004-0.008 W/kg. The blood-brain barrier envelopes the brain and protects it from toxic substances[4] DNA damage in cells exposed to RFR at SAR of 0.0024-0.024W/kg velizarov showed a decrease in cell proliferation (division) after exposure to RFR of 0.000021-0.0021W/kg. Biological effects do not necessarily mean adverse health effects. Some effects are reversible. If it is within certain limit. Similar research on human is yet to be done for further confirmation. WHO, on their website have shown that EMF protection measures have been taken by many countries, whereas status of Bangladesh is unknown. A comparison with china and USA is shown in Table 2: IV. CONCLUSION The intent of this paper is not to present an alarmist view of serious threat due to the radiations from the use of cellular phones, but to argue that even a small effect could create an epidemic size problem. Hence there is a risk from radiation of wireless devices. Bangladesh Telecommunication Regulatory Commission (BTRC) should take protection measures and monitoring, immediately to control radiation from licensed wireless equipments. As per standards defined by recognized bodies such as ITU, WHO, IEEE, FCC etc.
V. REFERENCES
[01] [02] [03] [04] [05] [06] [07] [08] [09] [10] [11] [12] [13]
American National Standards Institute (ANSI), Recommended Practice for the Measurement of Potentially Hazardous Electromagnetic Fields RF and Microwave. ANSI/IEEE C95.3-1992 James C. Lin, Health aspects of wireless communication: blood-brain barrier, cancer, and cell phone radiation, ACM SIGMOBILE Mobile Computing and Communications Review January 2002 American National Standards Institute (ANSI), Safety Levels with Respect to Human Exposure to Radio Frequency Electromagnetic Fields, 3 kHz to 300 GHz, ANSI/IEEE C95.1-1992. Jain, R., Wullert J., Challenges: Environmental Design for Pervasive Computing Systems, MOBICOM02, September 23-28, 2002, Atlanta, Georgia, USA. James C. Lin, Health aspects of wireless communication: Health and safety associated with exposure to wireless radiation from personal telecommunication base stations, ACM SIGMOBILE Mobile Computing and Communications Review June 2002 Volume 6 Issue 3. Yilong Lu, Mobile Phones and Health, http://www.ntu.edu.sg/home/eylu/fdtd/phone0.htm Foster KR, Vecchia P., Moulder JE., Health Effects of Mobile Phones: Recent Scientific and Policy Developments, http://www.seas, upenn.edu/~kfoster/COST259.htm Lai, Henry, Biological Effects of Radio Frequency Radiation from Wireless Transmission Towers, http://www.ecectric- words.com/cell/research/laisingh/celltower.htm. WHO Fact Sheets are at the WHO EMF website at http://www.who.int/peh-emf/ IEEE standards for Safety Levels with Respect to Human Exposure: Radio Frequency Electromagnetic Fields 3 kHz to 300 GHz, IEEE C95, 1991, Apr.1992. National Council on Radiation Protection and Measurements (NCRP), A Practical Guide to the Determination of Human Exposure to Radiofrequency Fields, NCRP Report No.119, 1993. James C. Lin, Health aspects of wireless communication: Cell phone testing and fundamental scientific research, ACM SIGMOBILE Mobile Computing and Communications Review January 2002.

65
Md. Aman Mahbub et al.
Robertson-Walker Metric and Cosmology

Md. Aman Mahbub1 and Md. Showkat Ali2
ABSTRACT
This paper describes how coordinates can be chosen so that the line element of a homogenous, isotropic space-time has the general Robertson-Walker form. A space-time is homogeneous and isotropic when each of its points lies on some member of a family of space like three-surfaces whose three-dimensional geometry is homogeneous and isotropic. Robertson-Walker coordinates employ one coordinate t to label the surfaces and three coordinates xi to label points in them. This constitutes a particular way of dividing space-time up into space and time. A particle of the cosmological fluid at xi on the surface is considered. Then using the same three coordinates an attempt was made to label the location of that particle on all other surfaces, i.e. for all time. In this paper, we have discussed about the cosmology, transformation of coordinates and derive the Robertson-Walker line element. From that a calculation was made to find out the Friedmann three models flat, open and closed. Keywords: Robertson-Walker Metric, consmology, red shight.
I. INTRODUCTION Professor Dr Robertson had a remarkable ability for getting to the crux of a problem and presenting his conclusions in such a manner that every one can understand that easily. Robertson makes an outstanding contribution to applied mathematics especially in the fields of differential geometry, quantum theory, the theory of general relativity and cosmology. His contributions to differential geometry came in papers Title: The absolute differential calculus of a non-Pythagorean nonRiemannian space in 1924; Transformation of Einstein space in 1925; Dynamical space-times which contain a conformal Euclidean 3-space in 1927. Applying his mathematical skills to cosmology had been an interest of Robertson's from early in his career as the paper on relativistic cosmology in 1928. He built on de Sitter's solution of the equations of general relativity in an empty universe and developed what are now called Robertson-Walker spaces. In 1928 and 1929, Robertson developed fully the "postulate of uniformity" so as to obtain the complete family of line-elements from the theory of continuous groups in Riemannian space. These Robertson-Walker cosmological spaces are still fundamental; A G Walker rediscovered them in 1936. He wrote a masterly paper Relativistic cosmology in 1933 which summarized his own contributions to the subject up to that time and well as contributions of others. In three papers entitled Kinematics and world structure published in 1935 and 1936 Robertson looked at Milne's theory and Einstein's theory of gravitation and their respective application to cosmology, arguing strongly that Milne's objections to Einstein's theory are not valid. The Poynting-Robertson effect is so named for Robertson's paper Dynamical effects of radiation in the solar system in which he studied the behaviour of a moving body, subject to the laws of general relativity, which is absorbing and re-emitting solar radiation. As a final illustration of Robertson's interest in the connection between geometry and physics we quote A H Taub's review of Robertson's paper. The geometries of the thermal and gravitational fields [1] which was published in the American Mathematical Monthly in 1950:"a study of the physical geometry of the gravitational and of the thermal fields, and of the reasons for the success of the former as opposed to the latter, as a basis for a tenable physical theory" [2]. The discussion of gravitation is given in terms of a four-dimensional scalar theory [3] but the main ideas of Einstein's general theory are stressed. The results of this theory regarding the motion of test particles and the behaviour of light rays are stated [4] and compared to that of Einstein's theory and differ from that theory. II. THE FRIEDMANN-ROBERTSON-WALKER METRIC It is very simple problem if 4-dimensional time-space is considered. We can choose the same time coordinate for each point in space, and at each time slice, we must have the isotropic and homogeneous three dimensional metric [5-6]. Aside from isotropy and homogeneity, general relativity requires that locally (e.g. near the origin) the line element is invariant under Lorentz transformations. Thus we arrive at the Friedmann-Robertson-Walker (FRW) metric, which is the most general metric (up to coordinate transformations) fulfilling the cosmological principle [7]. However, that is evident if you think of how the geodesic equation is derived from the metric by the Euler-Lagrange equations: since the coefficient of dt2 is coordinate independent, there can be no terms proportional to in the geodesic equation.
66
Bluetooth Security And Its Weaknesses: An Analytical Study
III. THE RED SHIFT OF LIGHT The light gets red-shifted in the FRW universe due to the time dependent scale factor. Suppose there is a galaxy at rest in the co-moving coordinates, with radial coordinate r1, and we are at the origin. Light is emitted by this galaxy at time r1 with frequency t1 and received on earth at time t0 with frequency is 0. We can determine the relation between 0 and 1. Light travels on `null geodesics that is, on geodesics with ds = 0. Note that in at space where ds2 = dt2 / dx2, ds = 0 simply means that the speed of light is c = 1. However, it is possible to choose locally at coordinates about any point in curved space, so having light travel on null geodesics means that any observer along its path sees it go by with velocity c = 1. This shows that along a radial light path is in the FRW universe. The red shift only depends on the ratio of the scale factor at reception to the scale factor at emission. It is not simply a function of the relative motion of the source at the time of emission; in fact if the universe were to contract for a while, and then expand, it is possible for a source which is approaching us to emit light which we see as shifted to the red. It is useful to consider in terms of the wavelength instead of the frequency. We know that the wavelength of light just contracts and stretches with the scale factor, and that fact explains the whole red shift phenomenon. In this view it is not much like the Doppler shift. Suppose we view a thermal source with temperature T1 with absolute luminosity L. The power emitted per unit area per unit frequency is looks rather strange; for example, with k = r1 = 1, this distance diverges, even though we know that the k = 1 space is compact, and that the greatest [proper distance one can get from any other point in that space. However, the above result makes sense. Recall that for k = 1, the spatial slices of our space are 3-spheres. If the star is at the North Pole and we are observing from two points on the equator, then our telescope always points to north. It follows that ds = 1 even though the star is a proper distance away. The ratio of this squared distance to the squared distance between the same pair of particles at different time t must be independent of x. Different ratios at different locations would violate the assumption of homogeneity. IV. DERIVATION OF ROBERTSON-WALKER METRIC The universe appears to be homogeneous and isotropic [5-6] on a sufficiently large scale. It is unlikely that we are in a special position in the universe. This leads us to the assumption of the cosmological principle which states that, the universe looks like the same pattern from every position in space at a particular time and that all direction in space at any point are equivalent. If we consider a series of nonintersecting space-like hyper-surface, then all galaxies lie on the hyper-surface in such a way that the surface of simultaneity of the local Lorentz Frame at any galaxy coincides locally with the hypersurface. The galaxy velocities are orthogonal to the space-like hyper-surface. This hyper-surface can be labeled by a parameter which may be taking as the properties of any galaxy. Hence each hypersurface defines a moment of time. Now let t denotes the synchronized propertime of all galaxies and introduce coordinates ( x , x , x ) = constant for all galaxies. Then the spacetime metric can be written as
1 2 3
d s 2 = dt 2 hij dx i dx j ; i, j= 1,2,3 where hij = hij t , x1, x 2 , x 3

2
(1) constant at
The spatial distance d of any two nearby galaxies on the same hypersurface
(x +x , x + x , x + x ) is
1 1 2 2 3 3
( x 1, x 2 ,x 3 ) and
d 2 = hij xi x j
(2)
Let us consider a triangle formed by three nearby galaxies at constant specific time and the other galaxies at some time later. Then by the Cosmological principle the second triangle must be similar to the first and the magnification factor must be independent of the position of the triangle in the 3-space. This means that time can enter d
2
R (t ) in order that the ratios of small distances may be the same at all times. Thus
2
only that a common factor
ds 2 = dt 2 R 2 (t ) rij dxi dx j , i, j= 1,2,3 where rij = rij x1, x 2, x3 .(3)
three spaces
d = rij dx dx are homogeneous, isotropic The and independent of time. Hence this must be a space of
2 i j ( 3)
constant curvature. The 3-dimensional curvature tensor the symmetry properties of where k is constant.

R ij k l of such space can be expressed in terms of rij alone. From

( 3)
( 3)
R ij k l it can be expressed as
Rij k l =k (rik rjl ril rjk )
(4)
67
Md Shafkat Amin et al.
Let the 3-dimensional curvature becomes: ds = dt

2 2
( 3)
R ij k l of the space d 2 = rij dxi dx j has the from of ri j , Then the space-time metric
2
R 2 (t ) (dx 2 + dy 2 + dz 2 )
k 2 2 2 1 + ( x + y + z ) 4
(5) Using the transformation:
x = R sin cos , y = R sin sin , z = R cos , we can write equation (5) in the form dR2 + R2 (d 2 + sin 2 d 2 ) (6) ds 2 = dt 2 R 2 (t ) k 2 (1 + ) 2 4 R Again by the transformation
r=
Equation (6) can be written as
R k 1 + R 2 4
dr 2 (7) + r 2 d 2 + sin 2 d 2 ds 2 = dt 2 R 2 (t ) 2 1 k r
This is the Robertson-Walker metric. Now without loss of generality, let us study this Robertson-Walker metric using the value of k =1, 0, -1. When k =1, this has the position curvature
d 2 = R 2 (t ) d 2 + sin 2 d 2 + sin 2 d 2 Such a space can be embedded in a 4-dimensional Euclidean space as follows: w =R cos , z =R sin cos , x =R sin sin cos , w 2 + x 2 + y 2 + z 2 = R 2 in Using,. This corresponds to the hyper sphere y = R sin sin sin the Euclidean space
1 , R2
with
the
transformation
r = sin
then
the
metric
becomes
)}
ds 2 = dw 2 + dx 2 + dy 2 + dz 2 =
= constant, we get 2sphere x + y + z =R sin . It surface are = 4 R sin . As ranges from 0 to , as objects moves from the north 2 2 pole of the hypersurface to the opposite pole through successive 2-sphere of areas 4 R sin . The area increases until = /2 after which it decrease until it is zero at = . This behavior is similar to what happens on a 2-sphere in a three
2 2 2 2 2 2 2
R 2 d 2 + sin d 2 + sin 2 d 2 This shows that all points and directions on this 3-sphere are equivalent. Putting
)}
(8)
dimensional space if one goes from North Pole to the South Pole. The entire hyper surface is swept by the co ordinates ranges: 0 , 0 , 0 2 . The total volume is
v =
( 3)
g d 3
= (R d ) (R sin d ) (R sin sin d )
(9) = 2 2 R 3 When k = 0, the universe has zero spatial curvature. In this case
d 2 = R 2 d 2 + 2 d 2 + sin 2 d 2 (10) This is the ordinary Euclidean 3-space. The range of (, , ) is 0 , 0 , 0 2 . Therefore the spatial volume is infinite.

)}
68
Bluetooth Security And Its Weaknesses: An Analytical Study
When k = -1, the universe has negative spatial curvature. In this case the spatial part of the Robertson-Walker space time in a 3-space
2
which
2 2
corresponds
2 2
to
the
surface
( x 2 + y 2 +z 2) +w2 = R 2
in
the
Minkowski
space
ds = dw dx dy dz with r = sinh and d 2 = R 2 {d 2 + sinh 2 (d 2 + sin 2 d 2 )} (11) The range of (, , ) is 0 , 0 , 0 2 . So the spatial volume is infinite.
Einstein equations are
1 R = 8 G T g T + g (12) 2 Where c = 2, G is the gravitational constant, T is the energy-momentum tensor, is
&& 3 R = 4 G ( + 3 p )R + R (13) Where is the density of the mass energy, p is the pressure, where as space-like components gives && & R R + 2 R 2 + 2k = 4 G ( p ) R 2 + R 2 (14)
V. DISEUSSIONS
the cosmological constant. For the Robertson-Walker metric time-time component of equation (12) gives
Equation (13) and equation (14) constitute the Einstein field equation for the Robertson-Walker space-time.
The Robertson-Walker line element describes anon-static but isotropic and homogeneous universe. It is kinematical in nature and ignores the usual dynamical equations general relativity. Thus we see that the universe is completely empty yet it predicts the observed recession of nebulae. Thus this equation represents the condition for cosmological model.
VI. REFERENCES
[1.] [2.] [3.] [4.] [5.] [6.] [7.]
Foster, J., & Nightingale, J.D, 1979. A short course in General relativity, Longman Group Limited. Islam, J.N., 1992. An introduction to Mathematical Cosmology, Cambridge University Press Ltd. Schutz, B.F., 1985. A first course in General Relativity, Cambridge University Press. Tourrence P, 1992. Relativity and Gravitation, Cambridge University Press. Hawking, S.W., & Ellis, G.F.R., 1980. The large scale structure of space-time. Cambridge University Press Ltd. Nieuwenhout F., 1997. The Friedmann Model for the Expansion of the Universe. Internet. Bergman, P.G., 1992. An Introduction to the Theory of Relativity.
Bluetooth Security and its Weaknesses: An Analytical Study

Mohammad Shafkat Amin1, Fahima Amin Bhuyan2 and Mohammad Shyiq Amin3
ABSTRACT
Bluetooth uses radio waves to transfer information and hence it is very susceptible to attacks. This paper first investigates the Bluetooth system and security issues in Bluetooth networks, and then it concentrates on specific security measures in Bluetooth, mainly authentication, encryption, key management and ad hoc aspects. The flaws and possible security holes in the Bluetooth Security Specification are also inspected in order to identify the reasons for these security holes. Keywords: Bluetooth, LMP, Network security, PIN and Encryption.

69
Md Shafkat Amin et al.
I. INTRODUCTION
Cables have been used to connect computers to each other and to special accessories from the very early ages of the computer era. Security measures have been developed to secure these cable connections so that information can travel safely. Bluetooth is one of the solutions to form a cable-free environment. This paper concentrates on the security measures of Bluetooth and performs an in-depth study as to whether security measures in Bluetooth are sufficient enough, so that Bluetooth can be used for everyday communications. The rest of this paper is organized as follows. First, in section 2, an insight into Bluetooth technology is taken, what possible uses it has and how it is implemented. Next, in section 3, the security of distributed systems and ad hoc networks in general is considered. Then the Bluetooth security architecture and its weaknesses are studied in detail.
II. BLUETOOTH TECHNICAL SPECIFICATIONS
Bluetooth devices are categorized into three different classes by the power they use. A class 3 device has a 1 mW transmission power and a range of 0.1-10 meters. A class 2 device has a transmission power of 1-2.5 mW and a 10-meter range. A class 1 device has a transmission power up to 100 mW and a range up to 100 meters. The architecture of Bluetooth is formed by the radio, the base frequency part and the Link Manager. Bluetooth uses the radio range of 2.45 GHz. The theoretical maximum bandwidth is 1 Mb/s, which is slowed down a bit by Forward Error Correction (FEC). Bluetooth specification designates the frequency hopping to be implemented with Gaussian Frequency Shift Keying (GFSK). The base frequency part of the Bluetooth architecture uses a combination of circuit and packet switching technologies. Bluetooth can support either one asynchronous data channel and up to three simultaneous synchronous speech channels, or one channel that transfers synchronous data and synchronous speech simultaneously. The Link Manager is an essential part of the Bluetooth architecture. It uses Link Manager Protocol (LMP) to configure, authenticate and handle the connections between Bluetooth devices. It also operates the power management scheme, which is divided into three modes: sniff, hold and park [1]. Several Bluetooth devices can form an ad hoc network. In these piconets, one of the Bluetooth devices will act as a master and the others are slaves. The master sets the frequency-hopping behavior of the piconet. It is also possible to connect up to 10 piconets to each other to form so called scatternets.
III. BLUETOOTH SECURITY
In every Bluetooth device, there are four entities used for maintaining the security at the link level. The Bluetooth device address (BD_ADDR), which is a 48-bit address that is unique for each Bluetooth device and defined by the Institute of Electrical and Electronics Engineers (IEEE). Private authentication key, which is a 128-bit random number used for authentication purposes. Private encryption key, 8-128 bits in length that is used for encryption and a random number (RAND), which is a frequently changing 128-bit random or pseudo-random number that is made by the Bluetooth device itself [2]. In Bluetooth Generic Access Profile, the Bluetooth security is divided into three modes: Security Mode 1: non-secure Security Mode 2: service level enforced security Security Mode 3: link level enforced security The difference between Security Mode 2 and Security Mode 3 is that in Security Mode 3 the Bluetooth device initiates security procedures before the channel is established [3]. There are also different security levels for devices and services. For devices, there are 2 levels, "trusted device" and "untrusted device". The trusted device obviously has unrestricted access to all services. For services, 3 security levels are defined: services that require authorization and authentication, services that require authentication only and services that are open to all devices [3]. The general architecture is shown in Figure 1. The key component is a security manager with the following tasks: Store security-related information on services. Store security-related information on devices. Answer access requests by protocol implementations or applications (access granted or refused).
1 Key Management
All security transactions between two or more parties are handled by the link key. The link key is a 128-bit random number. It is used in the authentication process and as a parameter when deriving the encryption key. The

70
Md. Mahbubul Alam Joarder et al.
lifetime of a link key depends on whether it is a semi-permanent or a temporary key. A semi- permanent key can be used after the current session is over to authenticate Figure 1: Bluetooth Security Architecture Enforce authentication and/or encryption before connecting to the application. Initiate or process input from an ESCE1 (e.g., the device user) to set-up trusted relationships on device level. Initiate pairing and query PIN entry by the user. PIN entry might also be done by an application. Bluetooth units that share it. A temporary key lasts only until the current session is terminated and it cannot be reused. Temporary keys are commonly used in point-to-multipoint connections, where the same information is transmitted to several recipients. There are several different types of keys defined in Bluetooth. Link keys can be combination keys, unit keys, master keys or initialization keys, depending on the type of application. In addition to link keys, there is the encryption key.
Figure 2: Different link keys between devices The unit key is generated in a single device when it is installed. The combination key is derived from information from two devices and it is generated for each new pair of Bluetooth devices. The master key is a temporary key, which replaces the current link key. It can be used when the master unit wants to transmit information to more than one recipient. The initialization key is used as link key during the initialization process when there are not yet any unit or combination keys. It is used only during the installation. This is depicted in figure 2. The length of the Personal Identification Number (PIN) code used in Bluetooth devices can vary between 1 and 16 octets. The regular 4-digit code is sufficient for some applications, but higher security applications may need longer codes. The PIN code of the device can be fixed, so that it needs to be entered only to the device wishing to connect. Another possibility is that the PIN code must be entered to the both devices during the initialization. The initialization key is needed when two devices with no prior engagements need to communicate. During the initialization process, the PIN code is entered to both devices. The initialization key itself is generated by the E22 algorithm, which uses the PIN code, the Bluetooth Device Address of the claimant device and a 128-bit random number generated by the verifier device as inputs. The resulting 128-bit initialization key is used for key exchange during the generation of a link key. After the key exchange the initialization key is discarded. The Key generation algorithm E22 is depicted in Figure 3.
Figure 3: Key generating algorithm E22 for master and initialization keys The unit key is generated with the key generating algorithm E21 when the Bluetooth device is in operation for the first
USB Token Authentication Technology for Secure Computer System
time. This is depicted in Figure 4. After it has been created, it will be stored in the non-volatile memory of the device and is rarely changed. Another device can use the other device's unit key as a link key between these devices. During the initialization process, the application decides which party should provide its unit key as the link key. If one of the devices is of restricted memory capabilities (i.e. cannot remember any extra keys), its link key is to be used.
Figure 4: Key generating algorithm E21 for unit and combination keys The combination key is generated during the initialization process if the devices have decided to use one. It is generated by both devices at the same time. First, both of the units generate a random number. With the key generating algorithm E21, both devices generate a key, combining the random number and their Bluetooth device addresses. After that, the devices exchange securely their random numbers and calculate the combination key to be used between them. The master key is the only temporary key of the link keys described above. It is generated by the master device by using the key generating algorithm E22 with two 128-bit random numbers. As all the link keys are 128 bits in length, the output of the E22 algorithm is 128 bits, too. The reason for using the key generating algorithm in the first place is just to make sure the resulting random number is random enough. A third random number is then transmitted to the slave and with the key generating algorithm and the current link key an overlay is computed by both the master and the slave. The new link key (the master key) is then sent to the slave, bitwise XORed with the overlay. With this, the slave can calculate the master key. This procedure must be performed with each slave the master wants to use the master key with.
Figure 5: Key generating algorithm E3 for the encryption key The encryption key is generated from the current link key, a 96-bit Ciphering Offset Number (COF) and a 128-bit random number. The COF is based on the Authenticated Ciphering Offset (ACO), which is generated during the authentication process. This process is depicted in Figure 5. When the Link Manager (LM) activates the encryption, the encryption key is generated. It is automatically changed every time the Bluetooth device enters the encryption mode.
2 Encryption
The Bluetooth encryption system encrypts the payloads of the packets. This is done with a stream cipher E0, which is re-synchronized for every payload. The E0 stream cipher consists of the payload key generator, the key stream generator and the encryption/decryption part [4]. The payload key generator combines the input bits in an appropriate order and shifts them to the four Linear Feedback Shift Registers (LSFR) of the key stream generator. The key stream bits are generated by a method derived from the summation
stream cipher generator [5]. Depending on whether a device uses a semi-permanent link key or a master key, there are several encryption modes available. If a unit key or a combination key is used, broadcast traffic is not encrypted. Individually addressed traffic can be either encrypted or not. If a master key is used, there are three possible modes. In encryption mode 1, nothing is encrypted. In encryption mode 2, broadcast traffic is not encrypted, but the individually addressed traffic is encrypted with the master key. And in encryption mode 3, all traffic is encrypted with the master key.
Figure 6: Description of the encryption process As the encryption key size varies from 8 bits to 128 bits, the size of the encryption key used between two devices must be negotiated. In each device, there is a parameter defining the maximum allowed key length. In the key size negotiation, the master sends its suggestion for the encryption key size to the slave. The slave can either accept and acknowledge it, or send another suggestion. This is continued, until a consensus is reached or one of the devices aborts the negotiation. The abortion of the negotiation is done by the used application. In every application, there is defined a minimum acceptable key size, and if the requirement is not met by either of the participants, the application aborts the negotiation and the encryption cannot be used. This is necessary to avoid the situation where a malicious device forces the encryption to be low in order to do some harm. The encryption process is depicted in Figure 6. The encryption algorithm uses four LFSRs of lengths 25, 31, 33 and 39, with the total length of 128. The initial 128-bit value of the four LFSRs is derived from the key stream generator itself using the encryption key, a 128-bit random number, the Bluetooth device address of the device and the 26-bit value of the master clock. The feedback polynomials used by the LFSRs [5] are all primitive, with the Hamming weight of 5. The polynomials used are (25, 20, 12, 8, 0), (31, 24, 16, 12, 0), (33, 28, 24, 4, 0) and (39, 36, 28, 4, 0).
3. Authentication
The Bluetooth authentication scheme uses a challenge-response strategy, where a 2-move protocol is used to check whether the other party knows the secret key. The protocol uses symmetric keys, so a successful authentication is based on the fact that both participants share the same key. As a side product, the Authenticated Ciphering Offset (ACO) is computed and stored in both devices and is used for cipher key generation later on. First, the verifier sends the claimant a random number to be authenticated. Then, both participants use the authentication function E1 with the random number, the claimants Bluetooth Device Address and the current link key to get a response. The claimant sends the response to the verifier, who then makes sure the responses match. The used application indicates who is to be authenticated. So the verifier may not necessarily be the master. Some of the applications require only one way authentication, so that only one party is authenticated. This it not always the case, as there could be a mutual authentication, where both parties are authenticated in turn.
Figure 7: Description of the authentication process If the authentication fails, there is a period of time that must pass until a new attempt at authentication can be made. The period of time doubles for each subsequent failed attempt from the same address, until the maximum waiting time is reached. The waiting time decreases exponentially to a minimum when no failed authentication attempts are made 4. Ad Hoc Aspects
There are some aspects of Bluetooth security that should be considered in the light of ad hoc networking. In an ad hoc network formed in a conference room, there are a couple of possibilities for Bluetooth devices to secure the traffic. First of all, they can use the combination keys to encrypt the traffic. This means that the master device forms combination keys with every slave device in the network. Then the information from a slave is subsequently sent to all other slaves by the master. Another way of forming a secure ad hoc network is to use the master key concept. Then all the devices in the network can use the same key when encrypting the traffic and no separate relaying of traffic is needed [6]. This seems to be the limit of the ad hoc aspects of the link level security mechanisms of Bluetooth. If there is to be more complex ad hoc networking, the security must be done on the application level. For example, if any KDCs or distributed secret schemes are to be used, Bluetooth does not support them directly [7].
IV. PROBLEMS IN THE SECURITY OF BLUETOOTH
In the encryption scheme of Bluetooth there seems to be some weaknesses. The E0 stream cipher with 128-bit key length can be broken in O(2^64) in some circumstances [8]. There is a divide and- conquer type of attack that is possible to perform, if the length of the given keystream is longer than the period of the shortest LFSR user in the key stream generation in E0. This, however, has been taken into account in the Bluetooth specifications. The above mentioned divide-and-conquer attack needs access to the key stream extending over periods of partial input. This can never happen, as Bluetooth has a very high re-synchronization frequency, i.e. the key stream segment generated to encrypt each frame is independent and so short that no possibilities arise for attacks. There is a problem in the usability of the Bluetooth devices, too. The use of the PIN code in the initialization process of two Bluetooth devices is stringent. The PIN code has to be entered twice every time two devices are required to be connected, it gets annoying even with shorter codes. If there is an ad hoc network of Bluetooth devices and every machine is to be initialized separately, it is unbearable. And it does not make upholding the security very easy. The specification makes a suggestion to use application level key agreement software with the longer (up to 16 octets) PIN codes. So the PIN code need not be entered physically to each device of the connection, but is exchanged with, for example, Diffie-Hellman key agreement. The generation of the initialization key may also be of concern. The strength of the initialization key is based purely on the used PIN code. The E22 initialization key generation algorithm derives the key from the PIN code, the length of the PIN code and a random number, which is transmitted over the air. The output is highly questionable, as the only secret is the PIN code. When using 4 digit PIN codes there are only 10.000 different possibilities. Adding the fact that 50% of used PINs are "0000", the trustworthiness of the initialization key is quite low. There is also a problem in the unit key scheme. Authentication and encryption are based on the assumption that the link key is the participants' shared secret. All other information used in the procedures is public. Now, suppose that devices A and B use A's unit key as their link key. At the same time (or later on), device C may communicate with device A and use A's unit key as the link key. This means that device B, having obtained A's unit key earlier, can use the unit key with a faked device address to calculate the encryption key and therefore listen to the traffic. It can also authenticate itself to device A as device C and to device C as device A. The Bluetooth Device Address, which is unique to each and every Bluetooth device, introduces another problem. When
a connection is made that a certain Bluetooth device belongs to a certain person, it is easy to track and monitor the behavior of this person. Logs can be made on all Bluetooth transactions and privacy is violated. Profiling and other questionable ways of categorizing can take place. Yet another problem with Bluetooth is the battery draining denial of service scheme, against which it has no protection. All in all, there are several problems still in the security of Bluetooth. It seems to be adequate for smaller applications, but any sensitive or otherwise problematic data should not be transmitted via Bluetooth.
V. CONCLUSIONS
In this research paper, the security specification of the Bluetooth system was examined. As was seen, the Bluetooth's security seemed to be adequate only for small ad hoc networks, such as a network of the participants in a meeting. Connecting a PDA to a mobile phone using Bluetooth may also be secure enough. In the light of this study, it seems that the security of Bluetooth is still inadequate for any serious, security sensitive work. After the basic problems have been corrected, the more sophisticated security methods may be implemented on the upper levels. The security specification only considers simple issues and the more functional security has to be built above it. This includes the better authorization systems with possible KDCs and distributed secret schemes. The secure routing protocols for larger ad hoc networks must also be implemented separately.
REFERENCES
[01] [02] [03] [04] [05] [06] [07] [08] [09] K. Sand, Bluetooth, Seminar on Multimedia, Telecommunications Software and Multimedia Laboratory, Helsinki University of Technology, Helsinki University of Technology, Finland, March 1999. http://www.tml.hut.fi/Opinnot/Tik111.550/1999/Esitelmat/Bluetooth/bluetooth.html. The Bluetooth Specification, http://www.bluetooth.com /developer/specification/specification.asp Muller T., Bluetooth Security Architecture: Version 1.0., Bluetooth White Paper, Document # 1.C.116/1.0, July 15, 1999. Saarinen M-J, A Software Implementation of the BlueTooth Encryption Algorithm E0, 2000-03-08[referred2000-03-15]< http://www.jyu.fi/~mjos/e0.c > Schneier B., Applied Cryptography, 2nd Ed., John Wiley & Sons Inc., 1996, 758p. Stajano F. & Anderson R., The Resurrecting Duckling: Security Issues for Ad-hoc Wireless Networks,Security Protocols Workshop,1999 Zhou L. & Haas Z., Securing Ad Hoc Networks IEEE NETWORK,1999 Hermelin M. & Nyberg K., Correlation Properties of the Bluetooth Combiner Generator, Information Security and Cryptology, 1999. 2007 Journal of Computer Science Vol 1, No 1, June 2007, pp 71-77 Faculty of Computer Science ISSN 1994-6244

Md. Mahbubul Alam Joarder1, Taufiq Hasan Malik2,S.M.Khalilullah1, Ali Sabbir1
and Shareeful Islam1 ABSTRACT

Information technology brought out a revolutionary information era where life came to be data driven in major extent. Data being so life oriented recent world events have brought concerns over the security in public information areas to the forefront of society's attention. Rapid increment of dependence upon data forces the involvement to the challenge of developing a reinforced security system that can ensure unauthorized access to the vital data through out its own way. The importance of security and effectiveness of information technology in developing security systems make the use of digital security systems not only for data but also every area of life aspects. Many business and homes are now installing security and access control systems to protect their properties. The digital identification system is now becoming the essential part of the overall security system. Installing a reliable and secured access control system is often a costly exercise that is too expensive for most people. The need for a low cost access control solution inspired the work presented in this paper. This paper presents a model for low
cost and highly effective digital identification system based on USB token. Keywords: USB token, digital security, digital identification, USB token registration, login access control, program access control and automated security warning mail.
I. INTRODUCTION
An access control system is simply an application program that prevents unauthorized usage of protected resources. The application program can be executed in digital computers or fabricated into digital electronic equipments to make up a real life intelligent security system. Automated digital security systems came to be a popular means of protecting unauthorized access to vital resources due to its reliability, effectiveness, efficiency, configurability and ability to communicate and interchange operational messages and instructions to and from the resource owner remotely through network or Internet. Despite of mass popularity of digital security systems, lack of cost effectiveness of such systems prohibits their availability and usability for general-purpose security issues. Researches have been undertaken in order to find out and develop a comparatively low cost and effective security system that can be made suitable for universal security uses. Using commercially viable design specifications derived from review of technologies, USB token is considered a better choice for developing digital identification systems for access control utilizing the facility of universality of USB technology and its low price tag. This paper depicts the practical application of USB token by developing a complete security system based on USB token that can even encourage further discovery of advanced USB token-based security mechanism for better security solution in general. The application program developed and described in this paper has three major functionalities as follows, USB token registration Login access control Program access control Automated security warning mail The details of design specifications and development strategies will be described later in this paper.
II. UNIVERSAL SERIAL BUS AND USB TOKEN
The Universal Serial Bus (USB) is a high speed, high bandwidth communications protocol for peripheral devices. USB incorporates many mechanical, electrical and software related specifications that make it a complex protocol. Tradition devices have communicated with the personal computer using simpler and slower, serial communications methods. The USB specification is defined as an interface for the connection of peripheral devices, for example printers, scanners, keyboards and pointing devices, to a PC or a similar host computer. A possible application for USB is in the field of measurement and control. As the need for general-purpose sensor and actuator devices in laboratories, households, offices and industrial applications increases, designers are faced with the problem of interfacing these devices with a computer. USB is an ideal solution for this problem.
Introduction to USB For USB to be successful it is important that the USB devices are user-friendly through having a plug-and-play type functionality. The USB specification is an industry standard, which addresses this by giving a complete definition of the bus system from the electrical interface to higher protocol layers, guaranteeing the interoperability of all devices. The basis specification of USB is supplemented by specifications for certain frequently used types of device classes, for example a human interface device (HID) class for keyboards, mouse pointers etc. 1. Plug and play functionality was one of the initial drivers of the USB specification; ease of use by consumers and customers has always been one of its main objectives. The massive penetration of USB became a reality through the integration of the USB hardware into chip-sets on PC Motherboards. Today it is estimated that at least two USB connections are available in every PC. USB has several key features that make it the first choice for designers now, and will help to keep it as a dominating protocol in the future: The same bus system is suitable for all attached devices. Using hubs, up to 127 end devices can be attached to the bus. Plug and Play: USB devices can be attached to or removed from the PC during operation. Small devices can be supplied with energy via the bus.
Minkowski Space Time In Theory of Relativity
Table 1: USB specifications and transmission rates Several versions of the USB specification exist. In table 1 the varying transmission rates are defined for the low, medium and high-speed USB protocols. However, these rates cover the speed of the entire bus system. If several end devices are attached to the bus, these devices must share the existing bandwidth. The latest specification, 2.0, was complied in the year 2000 by the USB Implementers Forum (USB-IF). It is compatible with version 1.1 and contains the already established Low Speed and Full Speed classifications. Additionally release 2.0 introduces a High Speed device type, however USB 2.0 compliant devices are only just becoming available today. Apart from the transmission speed, the USB Specification also standardizes the electrical and mechanical characteristics of the device connection (figure 1). In figure 1 the signals on lines D+ and D- are differential signals with voltage levels of 0V and 3.3V. These two lines transmit the USB data. Additionally, an operating voltage of 5V is made available for low-end devices, which only draw a current of up to 100mA.
Figure 1: Cross section of a USB cable and different USB plugs Most 8-bit microcontrollers with integrated USB functionality use the Low Speed implementation. Typical Low Speed applications are PC peripherals such as keyboards, mice and similar input devices. Motorolas innovative MC68HC908JB8 is a good example of a microcontroller with integrated USB functionality 2. This component contains the HC08-CPU clocked with 6MHz and the Low Speed USB interface as well as 8K Flash memory. The on-chip programming and debugging interface makes this MCU ideal for use in Low Speed applications and the development of USB applications in the laboratory. For measurement and control applications typically only small data ranges are needed, which is easily accomplished by using Low Speed USB devices. A Low speed USB device ensures a realistic data rate, which can be compared roughly with a RS232 connection with 9600 Baud. However, while RS232 represents a point-to-point connection, USB uses a bus structure. Several USB devices can be attached to a PC, using it as a bus master. If the connections at the PC host are not sufficient, hubs can be inserted and several hubs can be cascaded in up to 5 levels. Each USB device is addressed by the host via a unique address consisting of 127 addresses. The communication between host and USB device can be explained by a model of different software layers and protocols (figure 2). Communication always takes place between the appropriate layers of the host and the device firmware. This diagram references USB 1.1, however a full detailed discussion of software communication exceeds the limits of this article. For more information about this, Kelms book, USB 1.1 3 offers a good introduction.
VID and PID: The Identity of the USB Device The Vendor ID (VID) designates the manufacturer of the equipment. The assignment of this 16-Bit ID takes place via the USB organization, so that a clear allocation from devices to manufacturers can always take place. The use of unauthorized fantasy IDs would, in the long run, cause dangerous conflicts of different devices on the bus. The Product ID (PID) is specified at the manufacturers own discretion, and is also a 16-Bit number. VID and PID together usually represent the unique device identifier and are used to find the suitable driver component. Members of the USB-IF (Implementers forum) automatically receive a Vendor ID, however the yearly membership costs about $2,500 USD. This could quickly become a cost consideration, particularly for smaller companies, where USB is not the primary focus. However, to bring USB specified devices to market USB-IF membership is required, as it is necessary to have a legal VID. In addition, restrictions exist with use of the official USB logo. This logo is to signal the official "safety" of a product to end-customers, which is extremely useful in marketing mass-produced products to consumers, especially in
79
Md. Showkat Ali et al.
the PC market.
Digital Identification Methods Digital methods rely on data to identify a person. After investigating five principle methods identified in preliminary research, we chose one that was actually implemented as a web service. In order to rule out technologies, we used a series of criteria. The best solution could have been a combination of methods. Our initial research dealt with the five following digital identification techniques Key fob: A hardware token with built-in AES (Advanced Encryption Standard) encryption generating a constantly changing number sequence (PIN) that is required, in addition to a password, for a user to identify himself. The user would have a physical token, probably an attachment to his or her keychain (hence the name Key Fob), that contains a small display showing their current PIN. The PIN is generated algorithmically every 60 seconds. RFID wristband: A wristband containing an embedded RFID (Radio Frequency Identification) chip that broadcasts encrypted identification information, used to confirm the users identity. Traditional keys: In certain circumstances it may be appropriate to lock IT systems with a physical key, requiring users to have a copy. This allows tight control over system access, especially if the user group for a particular system is small. USB Electronic Token: This device plugs directly into a computers USB port and conveys a users ID information to software already running on the system. These devices are the size of a key and could be kept on a users keychain for easy access (USB Token Authentication Device, n. d.). Smart Card: A plastic card with an embedded chip that contains various pieces of ID information; a smart card provides not only memory capacity but also processing power, which enables encryption of its data (Fadely, 1998).
The power of these techniques comes from the fact that in order for a system to authenticate a user of one of these hardware tokens, the user must prove that he/she can correctly answer two questions: What do you have? and What do you know? The first question is answered with the physical device, the second with a pin or password. The advantage gained through this property is that if the answer to either question is compromised, the security of the system has not been put in threat. If a password is hacked, it is useless when not used in conjunction with the token and, conversely, if a token is stolen, it is powerless without a password. Changes Visible to Linux Users Beyond those general issues, Linux users will notice a few changes if they poke around through what the kernel tells them about their hardware. "usbfs" and /proc/bus/usb/devices You may know "usbfs" through its original and somewhat confusing name, of "usbdevfs". The name changed because it was completely unrelated to "devfs". If you use "usbfs" (perhaps through scripts like usbtree or tools like usbview) you will notice a few minor changes. /: Bus 03.Port 1: Dev 1, Class=root_hub, Driver=ohci-hcd/2p, 12M /: Bus 02.Port 1: Dev 1, Class=root_hub, Driver=ohci-hcd/3p, 12M Port 1: Dev 2, If 0, Class=hub, Driver=hub/4p, 12M /: Bus 01.Port 1: Dev 1, Class=root_hub, Driver=ehci-hcd/5p, 480M Port 2: Dev 2, If 0, Class=hub, Driver=hub/4p, 480M Port 1: Dev 3, If 0, Class=>ifc, Driver=pegasus, 12M Port 4: Dev 8, If 0, Class=hub, Driver=hub/3p, 12M Port 1: Dev 9, If 0, Class=HID, Driver=hid, 12M And in PROC file system: /proc/bus/usb/devices T: Bus=01 Lev=01 Prnt=01 Port=01 Cnt=01 Dev#= 2 Spd=480 MxCh= 0 D: Ver= 2.00 Cls=00(>ifc ) Sub=00 Prot=00 MxPS=64 #Cfgs= 1 P: Vendor=05ab ProdID=0060 Rev= 2.10 S: Manufacturer=In-System Design S: Product=USB Storage Adapter S: SerialNumber=2201010A0247C8AB C:* #Ifs= 1 Cfg#= 2 Atr=c0 MxPwr= 98mA I: If#= 0 Alt= 0 #EPs= 3 Cls=08(stor.) Sub=06 Prot=50 Driver=usb-storage E: Ad=01(O) Atr=02(Bulk) MxPS= 512 Ivl=125us
80
E: Ad=82(I) Atr=02(Bulk) MxPS= 512 Ivl=0ms E: Ad=83(I) Atr=03(Int.) MxPS= 2 Ivl=32ms The serial number mentioned here is changed dynamically with the change of USB devices. So according to no. 4 digital identification technique we can use it for uniquely identifying a user. Software running in a daemon can trace it and make a decision to access permission or do some predefined work associated with the user. We have implemented this principal in our project.
Figure 2: USB Token plugs into the USB drive of a computer
III USB Token
USB Token offers a secure and unique serial number. The serial number is stored in the token and it cannot be downloaded on to another device. The hardware is also tamper proof. Smart keys are technologically similar to USB token but interface with computers using a USB port. USB Token deploys two factor authentication schemes in which the user has to have something (USB Token) in possession and know something (password). The token is with the client at all times and therefore stealing it is not easy. The token is protected by a serial number. If a notebook is stolen the key can be discovered, but a USB Token is more difficult to steal and it also has a unique serial number. Several companies including Rainbow Technologies offer them for commercial and government use. USB Token however only provides secure authentication they do not address the issue of confidentiality or integrity.
Table 2: Comparison of various authentication procedures
IV GROWTH OF ACCEPTABILITY: USB TOKEN
81
USB devices, being portable, easy to install and plug and play to use, are being well accepted and used by a wide number of computer professionals and users. The IDC market analysis shows a report that shipments of iPods and other MP3 flash memory music players will surge nearly 124 million units in 2009 4. All these USB devices can be used for generalpurpose data storage as well. According to a 2005 FBI Computer Crime Survey, 44% of organizations have reported network intrusions from within their own organizations. Technology analyst Gartner warns that portable devices containing a USB or FireWire connection are a serious new threat to businesses. In their report, Gartner named removable media devices as a significant security risk in the workplace and advised that these can be used both to download confidential data, and also to introduce a virus into the company network. Any user can come into the office, plug in a USB stick the size of the average keychain and take in/out over 32 GB of data. This poses a tremendous threat: Users can take confidential data or they can unknowingly introduce viruses, trojans, illegal software and more actions that can affect whole network and company severely. An administrator has no way to control these social attacks until now. This means that a malicious insider can use a single USB storage to covertly take out (i.e. steal) proprietary data and millions of financial, consumer or otherwise sensitive corporate records at one go. This is why the USB token-based identification system can protect unauthorized access of unexpected USB devices to computers in organizations and protect vital business data. But this is to clarify that only USB devices through their unique tokens cannot guarantee identification of perfect owner of the devices because the device itself can be theft easily. In an Identity Fraud Survey, 2006, it is depicted that in 2005, the costs and damages caused by identity theft reached $56.6 billion 4 . However, USB token has come to be a better choice to reinforce the existing security solution mechanism due to its extreme portability and low cost issues. Anyone can be able to represent his digital signature through USB token from any corner of the world. The future probability and public acceptability of USB token-based identification mechanism 5 can be illustrated through the following data collected from online poll presented in figure 3.
Soft Token Biometrics
Authentication methods
A Combination PKI Keyfob USB Token Smart Card
Online poll result of number of users vs authentication method Number of total voter: 1750 First vote: Friday, 01 July 2005, 14:43 Last vote: Thursday, 19 April 2007, 21:31
Username/Password
200
400
600
800
1000
1200
Number of users
Figure 3: Online poll result illustrating usage of various e-
authentication methods.
V. DESIGN SPECIFICATION
The application developed as a security agent model is aimed to facilitate by the advantages of USB token. The basic requirements of a USB token-based security system can be identified as follows,6-15 USB standard specification awareness Robustness in data and work flow Time efficiency
82
Layered design approach Central access control mechanism Network support for remote assistance Multitasking Security warning mechanism Event logging Higher degree of Configurability
VI. SOFTWARE IMPLEMENTATION
This section describes the implementation strategies of the software based on the design specification described in the last section. The workflow of the major modules of the system is described in the following sections.6-15
USB Token Registration
The system is able to register new USB tokens by detecting new USB devices when plugged in by extracting token from the USB controller and looking up the system database. Administrators can plug in new USB devices for registering it for authenticity and configure the system for the newly registered USB token. The procedure of new USB device detection is illustrated in figure 4.
Figure 4: Flowchart of authenticating a new USB device
Login Access Control
The responsibility of the login access controller is to provide an efficient and secure user identification service. It also provides a reliable service that is able to process the inputs from readers and output with the corresponding action. The login access control system is a micro module of the authentication server that acts a login access control server and runs asynchronously in the background. When a request is sent to the server along with the user USB token it looks up the database to check the authorization information entries corresponding to the given token. Based on the gathered information from the database it sends positive or
USB Token Manager Authentication Server (AS) Login Access
User
Access Req
Authentication Request
JournalofComputerScience,Volume1,Number1,June2007,ISSN:19946244 ACK
Token (+ve/-ve)
83
Request for Login Access
Figure 5: Login Server negative acknowledgement to the requester. Upon the acknowledgement sent from the server, the client allows the user to login into the system. For workflow of login access controller is shown in figure 5.
Process Access Control
The process control software provides user identification service as well as gives user permission to some predefined application as authorized by the administrator. Process access control is another vital micro module of the central authentication server of the implemented system that runs asynchronously in the background. On receiving request from the client it extracts the USB token from the request and communicates with the database to check permission entries corresponding to the given token. Based on the gathered information the process access controller sets permission to the requested process for the requesting user and sends positive or negative acknowledgement to the requester. The workflow of process access controller is shown in figure 6.
Proc (USB Input)
BUS (Type of Device)
Device Info
No
String Search
File (Devices) = Null
Yes
Generate Error
No
Yes
String = Serial Number
Goto Old DB
Old DB of USB Serial Number
Compare new serial number with old DB
No
New Serial Number=True
Yes
Store New Serial Number to Auth. DB
Exit
Figure 6 Flowchart of process control
84
Automated security warning mail
Security warning mail system is an essential part of the security system implemented. It runs in the background at the server layer as a mailing daemon and is another micro module of the authentication server. The logic or program access controller triggers the security warning mail system after certain number of attempts of illegal access trial as set by the administrator.
Figure 7: Mail Daemon On being triggered the mail system sends predefined mail to the administrator or system authority. The workflow of the security warning system is shown in figure 7. The security warning mail system can be used with any other authentication services that can be added to the security system according to requirement, for example, the security system can be used to monitor presence of a number of people during a certain period of time. Appropriate monitoring service module can be implemented to monitor the USB tokens requested for login access during a certain period of time and thus can identify people who have not requested any login in time and can trigger the mail daemon to send notification messages to the appropriate authority mentioning about the absent ones. The system overview is shown in the block diagram presented in figure 8.
Figure 8: System Overview
85
VII. CONCLUSION
The USB token-based security system is growing up with its potential to be effective digital certification of identification and its easy to use feature along with its portability and comparatively low cost. Though USB token cannot be the candidate of stand-alone solution for the choice of digital identification mechanism due to the possibility of device stealth but it can be a better choice for reinforcing the security system. The model of USB token-based security system that is implemented and represented in this paper thus provides a very effective and generally reachable security solution that can be extended or customized according to the requirement of any large organization in order to arrange itself with the new security technology in an affordable manner.
VIII. REFERENCES
[01] [02] [03] [04] [05] [06] [07] [08] [09] [10] [11] [12] [13] [14] [15]
Van Eck, W. Electromagnetic radiation from video display units: an eavesdropping risk? Leidschendam, The Netherlands, (1985) SafeNet, Curtailing the Piracy Epidemic, A Case for Hardware Security Key, white paper, 2005 SafeNet, Inc. Radcliff, D. Busted! (Online) Computer World. Dec. (1998). De Trompet 2960 1967 DD HEEMSKERK, ID Control Identity and Access Management Page 2 of 5, THE NETHERLANDS (EUROPE) Contu R. and Girard J. Put Security Policies in Place for Portable Storage Devices, Gartner. (2004) John Leyden. Gummi bears defeat fingerprint sensors. The Register. 16 May 2002. Jan Wolter. A Guide to Web Authentication Alternatives. Dec 1997. Cathy Bowen, Government Smart Card IDs: Lessons Learned, Card Technology Magazine, November (2002). Donald Davis (2002), Early Steps Toward Biometric Standards, Card Technology agazine, November. Fancher C. H., In your Pocket: Smartcards, IEEE Spectrum (February), pp. 47-53(1997). Guillou L. C., et al., The smart Card: A Standardized security Device Dedicated to Public cryptology, in G.J. Simmons (Ed.), Contemporary Crypto-logy. The Science of Information Integrity, IEEE Press, pp. 561-613(1992). Jain A, Bolle R. and Pankanti S. BIOMETRICS: Personal Identification in Networked Society. Kluwer Academic Publishers, 1999. Paul B., and Raymund E., The Smart Diskette A universal token and personal crypto-engine, Advances in Cryptology-CRYPTO89, Lecture Note in Comp. Sc. pp. 74-79, G.Goos and J. Hartmanis (Eds). (1989) Quisquater J-J. The adolescence of smart card, Future Generation Computer Systems, pp.13 -37. (1997) Urien P., Internet Card, a Smart card as a true Internet node, Computer Communication, 23, pp. 1655-1666(2000).
Minkowski Space Time in Theory of Relativity

Md. Showkat Ali1 and Md. Aman Mahbub2
ABSTRACT
In physics and mathematics, Minkowski space (or, Minkowski space time) is the mathematical setting in which Einsteins theory of special relativity is most conveniently formulated. In this setting the three normal dimensions of space are combined with a single dimension of time to form a four-dimensional manifold for representing a Spacetime. Minkowski space is a four dimensional real vector space equipped with a non degenerate, symmetric bilinear form. Elements of Minkowski space are called events of four vectors. Minkowski space is often denoted R1,3 to emphasize the signature, although it is denoted by M4 or simply M. It is perhaps the simplest example of a pseudo Riemannian manifold. In this paper we discover how we can find Minkowski space time in null space and thus use it to describe the universe. Keywords: Minkowski space, theory of relativity and Riemannian manifold.
86
I. INTRODUCTION
Minkowski, Hermann (1864 - 1909), developed the concept of the spacetime continuum as a consequence of special relativity. To the three dimension of space, he added the concept of a forth dimension, time. The concept developed from Albert Einstein relativity theory and become in turn, the framework for Einsteins 1916 general theory of relativity. The characteristic feature of Galileo's Spacetime was the set of horizontal slices representing "planes of simultaneity". On a given plane, all of its events are simultaneous. This is the notion of Absolute Time, in which all observers agree on the elapsed-time between two given events. In the particular case of "zero elapsed-time", all observers agree that the events on a given horizontal plane are simultaneous. Einstein's extension of the Principle of Relativity to all physical laws requires us to abandon Galileo's Space time. In its place, we have the Einstein-Minkowski Spacetime. Hermann Minkowski developed a new view of space and time and laid the mathematical foundation of the theory of relativity. By 1907 Minkowski realized that the work of Lorentz and Einstein could be best understood in a non-Euclidean space. He considered space and time, which were formerly thought to be independent, to be coupled together in a four-dimensional 'space-time continuum'. Minkowski worked out a four-dimensional treatment of electrodynamics. In a paper published in 1908 Minkowski reformulated Einstein's paper by introducing the four-dimensional (space-time) non-Euclidean geometry, a step which Einstein did not think much of at the time. But more important is the attitude or philosophy that Minkowski, Hilbert - with whom Minkowski worked for a few years - Felix Klein and Hermann Weyl pursued, namely, that purely mathematical considerations, including harmony and elegance of ideas, should dominate in embracing new physical facts. In this view Minkowski followed Poincar whose philosophy was that mathematical physics, as opposed to theoretical physics, can furnish new physical principles [11]. This philosophy would seem to be a carry-over (modified of course) from the Eighteenth Century view that the world is designed mathematically and hence that the world must obey principles and laws which mathematicians uncover, such as the principle of least action of Maupertuis, Lagrange and Hamilton. Einstein was a theoretical physicist and for him mathematics must be suited to the physics [10]. This space-time continuum provided a framework for all later mathematical work in relativity. These ideas were used by Einstein in developing the general theory of relativity. In fact Minkowski had a major influence on Einstein considered mathematics to be a mere tool in the service of physical intuition. With the help of the mathematician Herman Minkowski (who gave us the idea to think in terms of "Spacetime", not just space and time separately), Einstein proposed a new model for Spacetime to replace Galileo's Spacetime. "Henceforth, space by itself, and time by itself, are doomed to fade away into mere shadows, and only a kind of union of the two will preserve an independent reality."
II. MINKOWSKI SPACE TIME
A main motive behind this change was the influence of two prominent German mathematicians: David Hilbert and Hermann Minkowski. Time has certainly proved Minkowski correct [3]. Murkowskis original mathematical interests were in pure mathematics and he spent much of his time investigating quadratic forms and continued fractions. His most original achievement, however, was his 'geometry of numbers' which he initiated in 1890. Geometrie der Zahlen was first published in 1910 but the first 240 pages (of the 256) appeared as the first section in 1896. Geometrie der Zahlen was reprinted in 1953 by Chelsea, New York, and reprinted again in 1968. Minkowski published Diophantische Approximationen: Eine Einfhrung in die Zahlentheorie in 1907. It gave an elementary account of his work on the geometry of numbers and of its applications to the theories of Diophantine approximation and of algebraic numbers. Work on the geometry of numbers led on to work on convex bodies and to questions about packing problems, the ways in which figures of a given shape can be placed within another given figure. Minkowski developed a new view of space and time and laid the mathematical foundation of the theory of relativity. In 1905, Albert Einstein published "On the Electrodynamics of Moving Bodies". In it, he emphasized the physical experimental fact that electromagnetic phenomena do not appear to be dependent on the absolute motion of the laboratory, but just on the relative motion of the various components in the experiment. In other words, it appears that "by performing electromagnetic experiments, you cannot distinguish between absolute rest and uniform motion." However, the Maxwell Laws of Electromagnetism (which is asserted to be correct) requires that the speed of light be the same constant, c, for every inertial observer. But this (as we have seen) cannot be maintained in Galileo's Spacetime. Einstein realized this inconsistency and could have chosen either: Keep Maxwell's Laws of Electromagnetism, and abandons Galileo's Spacetime or, keep Galileo's Spacetime, and abandon the Maxwell Laws. Recall that Galileo's Spacetime embodies Galileo's Principle of Relativity and Newton's Mechanical Laws of Motion. To give up Galileo's Spacetime would mean that there was something wrong with Galileo's Relativity or with Newton's Mechanics or with both. Einstein felt so strongly about Galileo's Principle of Relativity (which applied to the mechanical laws) that he extended it to include electromagnetic and optical laws. In other words, "no physical experiment (mechanical, electromagnetic, optical---or any physical law whatsoever) can distinguish between a state of absolute rest and a state of constant velocity." According, to Minkowski, the external world is not Euclidean space of three dimension, i.e. it is not composed of points whose coordinates are (x, y, z), where x, y, z are real numbers; but it is composed of events whose coordinates are ( x1 , x 2 , x3 ) are space coordinate and the forth one x 4 involves
1. Faculty of Computer Science, IBAIS University, Dhaka-1209. 2. Dept. of Mathematics, Jahangirnagar University. The authors like to thank Mr. Shahidul Alam for his contribution.
Matiur Rahaman Mian et al.
time. If and event occurs in the space, then the position of the point where it occurs and the instant when it occurs both are represented by the location of the event in the forth dimensional continuum. All the four dimensions are not equivalent. For, an axis which measures the distance in x direction, can be rotated to measure the distance in y and z direction; but the same axis can not be rotated to measure the time interval. Therefore the direction of time interval is not unique. This distinction is expressed by saying that the space time continuum is 3+1 dimensional rather than as four dimensional. The Minkowski space time ( , ) is the simplest empty spacetime in general relativity, and is infecting the space-time of special relativity. Mathematically, it is the manifold R4 with a flat Lorentz metric . In terms of the natural coordinates
( x 1 , x 2 , x 3 , x 4 ) on R4, the metric can be expressed in the form
ds 2 = (dx 4 ) 2 + (dx1 ) 2 + (dx 2 ) 2 + (dx 3 ) 2 (1)

These coordinates must be independent of transformation from one system to another system. If we use spherical coordinates (t , r , , ) where
x 4 = t , x 3 = r cos , x 2 = r sin cos 1 3 4 and x = r sin sin , then on calculations, dx = dt , dx = r sin d + cos dr
2 1 dx = sin cos dr + r cos cos d r sin sin d dx = sin sin dr + r cos sin d + r sin cos d Substituting these values in metric (1), it reduces to
ds 2 = dt 2 + dr 2 + r 2 (d 2 + sin 2 d 2 ) ds 2 = dt 2 + dr 2 + r 2 d 2
where, d = d + sin
2
(2)
d 2 .
Nothing unusual will happen to the coordinates, but we will want to keep careful track of the ranges of the other two coordinates. In this case, of course we have < t < + , 0 r < + . Technically, the world line r = 0 represents a coordinate singularity and should be covered by a different patch, but we all know what is going on so we will just act like r = 0 is well behaved. Our task is made somewhat easier if we switch to null coordinates:
1 u = (t + r ) 2 (3) 1 v = (t r ) 2
With corresponding ranges given by,
< u<+,
< v<+ , v u
These ranges are as portrayed in the figure, on which each point represents a 2 sphere of radius r = u v.
Figure 1.1 The metric in these coordinates is given by,
1 1 du = dt + dt 2 2
84
Measurement of Sheet Resistivities using Four-Point Probe Method
1 1 dv = dt dt 2 2
then, u + v = t and u v = r.
du dv = dr
Therefore
ds 2 = 2 (du dv + dv du ) + (u v) 2 d 2
The absence in the metric of terms du , dv correspondents to the fact that the surface null.
2 2
(4) {u = constant}, {v = constant}are
Now we want to change the coordinate in which infinity takes on a finite coordinate value. A good choice is,
U = arc tanu , V = arc tanv

The ranges are now,
2 < U < + 2 , 2 < V < + 2 , V U
U = arctan(u)
2
U
2
Figure 1.2
Then, we have,
dU =
du 1+ u2 dv dV = 1+ v2
and
cos (arc tan u ) =

similarly
1 1+ u2
cos (arc tan v) =

Then we are led to,
1 1+ v2
du = sec 2 U dU ; dv = sec 2 V dV
du dv + dv du = 1 cos U cos 2 V
2
(dU dV + dV dU )
(5)
Therefore, the Minkowski metric in these coordinates is
85
2 ds 2 = (dU dV + dV dU ) 2 2 cos U cos V + (u v ) d

2 2
Again,
(6)
(u v ) 2 =
1 sin 2 (U V ) cos U cos 2 V

2
Therefore, the Minkowski metric finally takes form, 1 [ 2 (dU dV + dV dU )] ds 2 = 2 cos U cos 2 V
+ sin 2 (U V ) d 2 (7 )
This has a certain appeal, since the metric appears as a fairly simple expression multiplied by an overall factor. We can make it even better by transforming back to a time like coordinate and space like (radial) coordinate , via,
= U +V , = U V
with ranges, Then,
dU = 1 ( d + d ) 2
< < + , 0 < +

1 ( d d ) 2
dV =
Therefore,
dU dV =
1 2 (d d 2 ) 4
Therefore, the metric (7) takes the simple form,
= 2 1 where = (cos + cos ) 2
1 d 2 + d 2 + sin 2 d 2 cos U cos 2 V d 2 + d 2 + sin 2 d 2 (8) ds 2 =

2
The Minkowski metric may therefore be thought of as related by a conformal transformation to the unphysical metric,
ds 2 = d 2 + d 2 + sin 2 d 2 (9)
This describes the manifold S , where the 3 sphere is maximally symmetric a static. There is a curvature in this metric, and it is not a solution of the vacuum Einsteins equations. This should not bother us, since it is unphysical; the true physical metric, obtained by a conformal transformation, is simply flat space time. In fact the metric is that of the Einstein static universe, a static (but unstable) solution of Einsteins equations with a perfect fluid and a cosmological constant. Of course, the full range of coordinate on S3 would usually be < < +, 0 while Minkowski space is mapped into the subspace. The entire S3 can be drawn as a cylinder, in which each circle is a three- sphere, as shown below. The shaded region represents Minkowski space. We note that each point ( , ) on this cylinder is half of two- sphere, where the other half is the point, ( , ). We can unroll the shaded region to portray Minkowski space as a triangle, as shown in the figure 1.4. This is the Penrose diagram. Each point represents a two sphere. In fact Minkowski space is only the interior of the above diagram (including = 0 ); the boundaries are not part of the original space-time. Together they are referred to as conformal infinity. The structure of the Penrose diagram allows us to subdivide conformal infinity into a few different regions:
3
86
=0 =
=
Figure 1.3
i + = future timelike infinity ( = , = 0) i 0 = spatial infinity ( = 0, = ) i = past time-like infinity ( = , = 0)
j + = future null infinity ( = , 0 < < ) j = past null infinity ( = + , 0 < < )
( , t ) ( , r)
=
t = constant
j
i0
r = constant
j+
i+
Figure 1.4 ( j+ and j- are pronounced as seri plus and seri minus, respectively). We note that i+ , i0 and i- are actually points since = 0 and = are the north and south poles of s3. Meanwhile j+ and j- are actually null surfaces, with the topology of S2. There are a number of important features of the Penrose diagram for Minkowski spacetime. The points i+ and i- can be thought of as the limits of spacelike surfaces whose normal are timelike; conversely, i0 can be thought of as the limit of timelike surfaces whose normal are spacelike. Radial null geodesics are at 45 in the diagram. All time like geodesics begin at i- and end at i+; all null geodesics begin at j- and end j+; all spacelike geodesics both begin and end at i0. On the other hand, there can be null geodesic timelike curves that end at null infinity (if the becomes asymptotically null). III. CONCLUSIONS
0
Finally in this paper we have shown that it is easy to describe the universe using Minkowski space in null space. In this view we have shown that purely mathematical considerations including harmony and elegance of idea showed dominate in embracing new physical facts. Mathematics so to speak be master and physical theory would be made to bow to the master. Otherwise theoretical physics was a sub-domain of mathematical physics.
IV. REFERENCES
[01] [02] [03]
Dieudonne, J M Bliss, 1970-1990 Biography in Dictionary of Scientific Biography, New York. Biography in Encyclopedia Britannica. [Available on the Web] Hancock, H., 1939, Development of the Minkowski Geometry of Numbers ,New York.
87
[04] [05] [06] [07] [08] [09] [10]
Lanchester, F. W., 1935, Relativity: an elementary explanation of the space-time relations as established by Minkowski, and a discusson of gravitational theory based thereon, London. Benz,W.,Lorentz-Minkowski geometry,DeSitter's world and Einstein's cylinder universe, in Charlemagne and his heritage. 1200 years of civilization and science in Europe, Aachen, (Brepols, Turnhout, 1998). Corry, L., Hermann Minkowski and the postulate of relativity, Arch. Hist. Exact Sci. (1997). Corry, L, The influence of David Hilbert and Hermann Minkowski on Einstein's views over the interrelation between physics and mathematics. Kersten, S., Zu einigen weltans chaulich-philosophischen Ansichten Hermann Minkowskis,Wiss. Z. Pdagog. Hochsch. 'Karl Liebknecht' Potsdam. Serre, J-P., Smith, Minkowski et l'Acadmie des Sciences, Gaz. Math. (1993). Strobl, W., Aus den wissens-chaftlichen Anfngen Hermann Minkowskis, Historia Mathematica 12 (2) (1985), 142156.

Matiur Rahaman Mian1 and Md Osman Goni2
ABSTRACT
The sheet resistivities on n-dimensional rectangular and circular samples using the four-point probe method were measured. Diffused surface layers can be treated as two-dimensional structures, but the factors are also useful in obtaining body resistivities on thin samples. Keywards : Sheet resistivity, body resistivity, thin, thick, rectangular and circular sample
I. INTRODUCTION
The four point probe method has proved to be a convenient tool for the measurement of resistivities.
88
M. Mostafizur Rahman
This method was first described by L. Valdes [1], which gives the functional relationship between the resistivity and the voltage and current readings for various geometries. Later, A Uhlir [2-3] evaluated functions which give the relationship for additional geometries. All these treatments are concerned with three-dimensional structures infinite in at least one direction. For diffused layers, a similar relationship can be applied for the evaluation of sheet resistivities on various sample shapes. This is a two dimensional problem which is treated in this present works for various finite sample sizes. With the solution, however, it is also possible to find the body resistivities on thin sample slices of the same finite geometries.
Let us consider a rectangular sample with the dimensions a and d . Apply four- point probe method as shown in figure-1. The probe is arranged symmetrically with point spacing S . To obtain the voltage between the two center points marked 1 and 2, an infinite arrangement of dipoles must be considered, as shown in figure-2. All these contribute to the voltages between points marked 1 and 2. F. Ollendroff et al [5] gives the potential distribution for an infinite number of current sources, arranged in a line and equally spaced. With a coordinate system as in Figure-3 the potential is given by
2 Ip s 2 v v0 = l n 2 sin x + sinh y ..(5) d d 2
II. METHOD OF MEASUREMENT In the experiment, the renowned four-point probe method was used to determine the sheet resistivities on n-dimensional rectangular and circular samples. The basic concept of Four-Point Probe method is that if a current source is applied to an infinite sheet of samples, it gives rise to the logarithmic potential as
v v0 =
where v = the potential, I = the current, s = the sheet resistivity and r = the distance from the current source. In particular, the potential for a dipole (+ Source and Source) becomes
I s l n r .(1) 2
Figure-1: Arrangement of a four-point probe on a rectangular sample for the measurement of Sheet Resistivities
v v0 =
I s r l n 1 ..(2) 2 r2
In the case of a four-point probe on a sheet, the two outside current points represent the dipole. Therefore, the potential difference between the two inner points for an infinite sheet is given by
v = V =
I s
l n 2 ..(3)
Figure-2: System of images.
When only equal point spacing is considered, then the sheet resistivities of equation (3) becomes
s =
V V = .4.5321.... (4) I ln2 I
To obtain the sheet resistivity on a finite sample, the method of images can be applied [4]. Only nonconducting boundaries are considered; the boundary edges of a diffused layer must be etched or else the layer on the back side of the sample would act as a shunting path.
Figure-3 : Co-ordinate system for a linear arrangement of current-sources From expression (5), the present problem can be reduced to a summation of lines of current sources with alternating sign in only one direction. In the coordinate
III. FOR RECTANGULAR SAMPLES
Numerical Solution of one Dimensional Heat Conduction Equation by Finite Difference Method.
system for every line of sources the point marked 1 and 2 have the x-co-ordinate zero. Thus the expression simplifies as
I v v0 = s l n (e ry / d e ry / d ) (6) 2
( y +S) / d
Each line of sources thus contributes to the voltage V, the amount is given by
n I er e r ( yn + S ) / d ) (7) v n = s ln ( ry n / d ry m / d 2 e e where y n is the distance from point 1 to the center source in line n. The dimension a is involved in the values yn. The total voltage between the Points 1 and 2
1.00 1.25 1.50 1.75 2.00 2.50 3.00 4.00 5.00 7.50 10.0 15.0 20.0 40.0
2.266 2.929 3.363 3.927 4.172 4.365 4.436 4.508 4:532
2.457 3.114 3.509 4.009 4.221 4.388 4.452 4.512 4.532
1.479 1.720 1.945 2.353 2.700 3.225 3.575 4.036 4.236 4.397 4.455 4.513 4.532
0.999 1.247 1.489 1.724 1.947 2.354 2.701 3.225 3.575 4.036 4.236 4.395 4.455 4.513 4.533
0.999 1.225 1.489 1.724 1.947 2.354 2.701 3.225 3.575 4.030 4.236 4.395 4.455 4.543 4.532
is therefore
1 V = v n p s I ...(8) a d C ; d s where C (a / d ; d / s ) is constant. . As shown in the

calculation, the summation can be done easily by expanding the logarithm and summing each term as a geometrical series. In almost all cases, the first term gives four places accuracy. The sheet resistivitity is thus given by
Table- II: Calculated correction factor C for the Mearurement of Sheet Resistivities using Four-Point Probe on Narrow Structure specimens.
ps =
V a d C ; . .(9) I d s
Where ps =
The calculated correction factors C for different samples of various geometries are given in the table-1 and for very thin samples in table-II.. For small d / s , the quantity C = s / d C is very close to unity. In these cases the sheet resistivity can be expressed as:
vd C' I s Ratio a / d a/d d/s =2 =1

1.00 1.25 1.50 1.75 2.00 2.50 3.00 0.9859 0.9826 0.9727 0.9413 0.8192 0.9000
a/d
=3 0.9988 0.9973 0.9929 0.9850 0.9737 0.9416 0.9002
ps =
V d V d . ..(10) C' I s I s
0.9994 0.9974 0.9929 0.9850 0.9737 0.9416 0.9002
a/d 4
4.00
0.778 0.806 0.8062 4 1
0.8062
Table-1: Calculated correction Factor C for the Measurement of Sheet Resistivities using the Four-Point Probe Method.
The table II may be used to determine the error one makes by just approximating with C ' = 1 .
Where Ratio
d/s
v C I Circle a/d ps =
dia. d/S = 1
IV. FOR CIRCULAR SAMPLE In this case of circular sample only one image, it is necessary to fulfill the boundary conditions. The image is obtained by reflecting the dipole on the circle [5] For a four- point probe centered in the sample the voltage becomes:
a/d
=2
a/d
=3
a/d 4
2 d +3 Ip s ) V = x In 2 + In( 2 d 3 s 1 .(11) = Ips d C s where d is the diameter of the circle. The correction
= s .W =
V W .W . .F .(15) ln 2 S I F (a / v) w/ s
0.4000 0.5000 0.5555 0.6250 0.7143 0.8333 1.0000 1.1111 1.2500 1.4286 1.6666 2.0000 0.9995 0.9974 0.9948 0.9898 0.9798 0.9600 0.9244 0.8907 0.8490 0.7938 0.7225 0.6336
factor is also given in table I.
V. MEASUREMENT OF BODY RESISTIVITIES ON THIN SAMPLES

On a slice sample of thickness w, the four point probe is introduce voltage gradients perpendicular to the surface. Insofar as these gradients are negligible, the slice can be treated in the same way as an infinitely thin slice and the proper sheet resistivity can be obtained, thus giving the body resistivity p with the relation
p = ps w. ..(12)
As mentioned earlier, for a four- point probe method, on an infinite sheet the following relation holds:
s =
V V = 4.5321.... ..(4) I ln2 I
VI. CALCULATION Summation of v n [7]

Without the loss of generality the following substitutions are made:
But for an infinite slice of finite thickness w, one can
w V F express the resistivity as p = p s w = w I In 2 s

..(13) Where w / s is a correction factor approaching unity as w approaches to zero.
d Is used for
From Uhlirs paper [2] the term F w / s can be obtained using thick samples, tabulated in table - III. This table may be serving to evaluate the error one makes by treating thin samples as an infinitesimal sheet. For a finite sample, F w / s could be used as a first order connection, thus giving the resistivity of a slice
d a , a for , and s s y y is used for . s
From symmetry reasons, it follows that each source in the upper half of the plane (as shown in figure-2) contributes the same v n as does the corresponding source in the lower half. Thus only the sources in the lower half are to be considered and the result must be multiplied by 2. The terms l n (
p = p s w = (V / I ) w CF.(14)
The exact treatment would require the summation of the potential from an infinite three dimensional array of dipoles. Table- III : Measurement of Body Resistivities on thin Samples of thickness w .
ry / d ry / d ) can be written in e e
the form of
l n e ry / d (1 e 2 ry / d ) = y l n (1 e 2 ry / d ) .(16) d
From this equation, it is easy to find
2 r ( y +1) / d 2 v n = I n 1 e n I s d
+ I n 1 e 2 rynld ].(17)
Here + sign is used for a positive source and sign for a negative source. Summing the first term gives / d . To sum the l n terms the logarithm can be expanded as
l n (1 e 2 r / d ) + am ]
The first term in
m gives four-places of accuracy

VII. CONCLUSION
m =1
1 2 1 3 x + x + ... 2 3 with this, each term in n becomes 1 2 r ( yn +1)m / d e e 2 ryn m / d .(18) m m =1 Then y n can be expressed as y n = Ai + n2a with this each term in m becomes This is a geometrical series in n. n 4 1 4am / d 2Aim / d 2m / d am = e e e 1 m i =1 n = 0
l n (1 x ) = x +
The four-point probe method is very accurate for the measurement of sheet resistivities of different samples in different dimensions.
VIII. REFERENCES
[01] [02] Valdes L., Resistivity Measurements on Germanium for Transistors, Proe. I.R.E., 42, Feb., 1954, p.420. Uhlir, A.,Jr., The potentials of infinite Systems of Sources and Numerical Solutions of Problems in Semiconductor. Engineering, B.S.T.J.34, Jan., 1955, p.105
Using
4
n =0
we can write
[03]
am =
i =1
1 e 2 rm / d 1 2 rAim / d e m 1 e 4 ram / d
Miller E, Paces J and Komarek K L, Trans. Met. Soc. AIME, 230, 1557, 1964
Thomson Sir William, Electrostatics and Magnetism, Macmillan, London, 1884. Ollendorffi F., Potentialfelder der Elektrotechnik, Springer, Berlin. 1932. Encyclopadic der Mathematischen Wissenschaften, 5, 2. Teil, Teubner 2007 Journal of Computer Science B.G., Leipzig, 1904Vol 1, No 1, June 2007, pp 87-90 Faculty of Computer Science 1922, p. 422. ISSN 1994-6244 Smith E.M, The Bell system Technical Journal, p.711-718
[04] [05] [06]
For reasons of convergence, y =1 is treated separately. The Ai thus become For + sources: For - sources:
4
A2 = a + 1 A4 = 2a 2 A1 = 2a + 1 A3 = a 2
i 2 rAim / d
[07]
Forming
( 1) e
i =1
and inserting into
1 2 r (a 2 )m / d 1 e 6 rm / d (1 e 2 rm / d ) e m 1 + e 2 ram / d the expression for am gives am =
The total voltage between the points marked 1 and 2 is therefore
V = I s
1 [ + l n (1 e 4 r / d ) d
M. Mostafizur Rahman1
ABSTRACT
In this work the numerical solution by different finite difference method have been compared to find out the better methodology for solving one dimensional heat conduction partial differential equation. To find the batter feasible solution a comparative study between the exact solution of a parabolic heat conduction equation whose exact solution is known and its numerical solution provided by the different finite difference methods have been compared. To do so it is also make change of parameters to get different trial of feasibility of the methods. Keywords: partial differential equation, heat conduction equation and finite difference method
I. INTRODUCTION Generally in a partial differential equation, one independent variable is the time and the others are space co-ordinate; and the required function describes an unsteady process, more specifically a diffusion process. The basic example of a parabolic equation is the one dimensional heat conduction equation.
pU i 1, j +1 + (2 + 2 p) U i , j +1 pU i +1, j +1
= pU i 1, j + (2 2 p) U i , j + pU i +1, j
....(2)
Which is an implicit form, called Crank-Nicolson mtethod and is stable for all finite values of p. [7] Considering
2
III. EXPLICIT METHOD the heat conduction
equation
u 2u =K Here u (t , x ) represent the heat on a t x2

line at time t and position x and K is a constant called thermal conductivity[1]. In one space dimension k=1. So the equation becomes
u u = , let the (x, t) plane be divided into t x2

smaller rectangles by means of the sets of lines, x=ih, i=0, 1, 2.......... and t = jk, j=0, 1, 2....... using the approximation,
u 2u = ......................(1) t x2
In this work, the above equation is taken as test equation and its exact solution is known. Here it is solved to get numerical solution by Crank-Nicolson method, explicit method and fully implicit method and after that all the values are compared with the exact solution to put forward the batter methodology to solve heat conduction equation.
II. CRANK-NICOLSON METHOD
u u i , j +1 u i , j = ...................(3) k t 2 u u i 1, j 2u i , j + u i +1, j = ....................(4) x2 h2
Now equation (1) can be replaced by the finite difference analogue U i , j +1 U i , j U i 1, j 2U i , j + U i +1, j = k h2 Where U i, j is the finite difference approximation
to the exact solution u
i, j
Crank and Nicolson proposed a method in 1947 according to which
Which can be written as,
of its finite difference approximation on the jth and (j+1)th rows and known as Crank-Nicolson Method [2, 3] Then U i 1 , j 2 U i , j + U i + 1 , j 2 2 1 u h Hen = 2 2 U i 1 , j + 1 2 U i , j + 1 + U i + 1 , j + 1 x + h2
ce equation (1) can be replaced by
2u is replaced by the average x 2
i , j +1
=U
i, j
+ p (U i 1, j 2U i , j + U i +1, j )
= pU i 1, j + (1 2 p)U i , j + pU i +1, j where p = k h2
This formula express the unknown function value of the (j+1)th time levels in terms of the known function at jth level and hence it is called the explicit formula. This formula is valid only for 0 < p 1 [4] 2
IV. FULLY IMPLICIT METHOD
1 1 u i , j +1 u i , j = 2 k 2h
U i 1, j 2U i , j + U i +1, j + U i 1, j +1 2U i , j +1 + U i +1, j +1
which gives on rearranging
u 2u = A general approximation to is t x2
Md. Hanif Mian et al.
2u u = 2 x t
2u + (1 ) 2 x i , j +1
.......(5) i, j
where, 0 1 and
2u ( space derivative) is x2
replaced by a weighted average approximation of the values at time t j & t j +1 . This is called method [4] This method involves only two time levels, but other methods involving more than two time levels are possible [6]. Now substituting the equation (2) and (3) in the equation (5) we get,
numerical values of exact solution and numerical output by the above mentioned methods. We get three different tables of output. We also use here Techplot simulator to find out the graphical output of the data tables of three different methods. Due to the page limitation here only the graphical output is presented. Only a few data output is shown in a table by Crank-Nicolson method to show the convergence behavior of the solution. Figure 1 is the Graphical representation of the exact solution and numerical solution by the Explicit method at h=.05, k=.0005 at t =.5. Where both the graphs coincide with an accuracy up to four decimal.
0.007
U i , j +1 p( U i +1, j +1 2U i , j +1 + U
i 1, j +1
)
0.006
= U i , j + (1 ) p (U i +1, j 2U i , j + U i 1, j )
...(6) . This gives a set of implicit equation for values of U on the new time step if > 0 . For = 1 the fully implicit method is stable for all value of p. [5]
V. CONVERGENCE
____Numeric Value ---------Exact Value

(Sketch) 08 May 2007
______
Numeric Value
0.005
0.004
The finite difference equation is said to be convergent when u(x,t)-U(x,t), the difference between the solution of the differential and difference equations at a fixed point (x,t) tend to zero, as the mesh spacings tend to zero. In general the problem of convergence is a difficult one to investigate. However it is not necessary to establish convergence separately, it can be investigated in terms of consistency and stability [4] Lax's equivalence Theorem states that given a properly posed linear initial-value problem and a linear finite-difference approximation to it that satisfies the consistency condition, stability is the necessary and sufficient condition for convergence. [7]
VI. TEST PROBLEM AND RESULT
n
0.003 0.002 0.001 0 0
0.25
0.5 X
0.75
Figure 1: Graphical representation of the exact solution and numerical solution by explicit method at h=.05, k=.0005 at t=.5 Figure 2 is the Graphical representation of the exact solution and numerical solution by Crank-Nicolson method at h =.05, k=.0005 at =.5 with an accuracy up to three decimal. Where we see that there is a little gaze at the middle of the graph. ____Numeric Value ---------Exact Value
0.007
0.006
0.005
0.004
Consider the heat equation
With the boundary conditions and initial conditions
u ( x, t ) u 2 ( x, t ) = 0, t x 2
0 < x < 1,
0<t
u (0, t ) = u (1, t ) = 0 0 < t
n
0.003 0.002 0.001 0 0
0.25
0.5
0.75
u ( x,0) = sin(x ), 0 x 1
and the exact solution of the problem is
u ( x, t ) = e t sin( x)
Here we use FORTRAN programming language to simulate the methods of Explicit, Crank-Nicolson and the fully Implicit method and get different data set of 9 2
Figure 2: Graphical representation of the exact solution and the numerical solution by CrankNikolson method at h=.05, k=.0005 at t=.5 Table 1 is representing the error of the solution for h =.1, .05, .025 by Crank-Nicolson method with the fixed time step k =.0005 at t = .5. This will illustrate the convergence behavior of the solution in terms of order h, as h 0.
MS SQL Server in Web Browser
h x
Errors in different subintervals by Crank_Nicolson method h=.1 h=.05 h=.025
Figure 4: Graphical representation of the solution by Explicit, Implicit and Crank-Nikolson method with h=.05, k=.0005 at t=.5
VII. CONCLUSION
.1 .3 .5 .7 .9
-0.000091 -0.000240 -0.000297 -0.000240 -0.000091
-0.000022 -0.000058 -0.000072 -0.000058 -0.000022
-0.000005 -0.000014 -0.000018 -0.000014 -0.000005
Table 1: The error in the solution for h=.1, .05, .025 by Crank-Nicolson method at k=.0005 measured at t=.5 Figure 3 is the Graphical representation of the exact solution and numerical solution by the Implicit method at h=.05, k=.0005 at t=.5 with an accuracy up to three decimal. Where we see that there is aloofness between the graphs of the numerical result. ____Numeric Value ---------Exact Value
0.007 0.006
Going on process it need to be considerd small steps both in k and h and the results of all the methods can be improved by reducing the intervals. The explicit method requires very small k as compared with h for stability. So take long computation time. The fully implicit method is unconditionally stable but it requires small steps for accuracy. The CrankNicolson method gives reasonable results for moderated k and h. It is noted that rounding error has effects to compute solution of linear system. So these effects may cause some errors in the accuracy of the Crank-Nicolson method. So it may be concluded that to solve the heat conduction equation the CrankNicolson method is better for obtaining more accurate results.
VIII. REFERENCES
[01] [02] [03]

0 0.25 0.5 0.75 1
0.005
0.004
0.003
0.002
0.001
Figure: 3 Graphical representation of the exact solution and the numerical solution by implicit method at h=.05, k=.0005 at t=.5 In figure 4 we give the Graphical representation of the solution given by Explicit, Crank-Nicolson and fully implicit method at h=.05, k=.0005 at t=.5.
0.008
[04]
[05]
[06] [07]
0.007
0.006
0.005
0.004
0.003
[08]
0.002
0.001
0.25 0.5 0.75 1
[09]
Partial differential equations and finite difference methods in image processing--Part II: Image restoration Jain, A.; Jain, J. Automatic Control, IEEE Transactions on Volume 23, Issue 5, Oct 1978 Page(s): 817 834 Digital Object Identifier 2. S.S SHASTRY, Introductory Methods of Numerical Analysis, Third Edition, Prentice Hall of India, 1999. SCHAUM'S OUTLINE SERIES, Theory and Problems of Partial Differential Equations, First Edition, McGraw-Hill Book Company, 1986. G.D. SMITH, Numerical Solution of Partial Differential Equations, Finite Difference Methods, Second Edition, Clarendon Press, Oxford, 1978. RAISINGH HANIA, Ordinary and Partial Differential Equation, Sixth Edition, S. Chand & Company, 2000. 6. MITCHELL A.R and D.G GRIFFITHS, The Finite Difference Method in Partial Differential Equations, Wiley And Sons, 1980. 7. Applications of stencil-adaptive finite difference method to incompressible viscous flows with curved boundary ; H. Dinga, C. Shua, , and Q.D. Caib a Department of Mechanical Engineering, National University of Singapore, Singapore 119260, Singapore
93
Md. Hanif Mian et al.
[10] [11]
b State Key Laboratory for Turbulence and Complex Systems, Peking University, Beijing 100871, China Received 25 October 2005; revised 28 April 2006; accepted 30 May 2006. Available online 5 October 2006.
MS SQL Server in Web Browser

Md. Hanif Mian1, Jugal Krishna Das2 and Muhammad Ruhul Amin Khandaker3
ABSTRACT
To satisfy the evercan use graphical consoles. In this Microsoft SQL Web architecture. We encourage the 2007 Journal of Computer Science Vol 1, No 1, June 2007, pp 91-93 Faculty of Computer Science ISSN 1994-6244 increasing necessity of remote access of databases we interfaces through web browsers rather than text paper, we discuss the media for Web-administrating in Data Administrator which actively uses .NET then discuss some of its advantages which may readers to use this tool.
Keywords: Index Terms- SQL Server, SCDB, Cassini and .NET I. INTRODUCTION
Although the professionals of data base count that a better media to interact with server is text consoles, but the producers of Subjective Controls of Data Base (SCDB) take their course to the way of growing the graphical interfaces for their products. The real steps on this path is called the using of web-browser as a graphical client, moreover Ideologies web-interfaces are based on the filling up forms, well assigned in the mechanism for interaction with SQL server.
II. WEB ADMINISTRATING IN MS SQL SERVER
Not so long ago, Microsoft SQL Server provided itself the media for Web-administrating in Microsoft SQL Web Data Administrator, and very soon the second version of this product was released. Microsoft SQL Web Data Administrator is distributed free of charge (downloadable in Microsoft site). Microsoft SQL Web Data Administrator 2.0 actively uses architecture .NET (Systems requirement of this product - Windows 2000 SP2 or XP, .NET Framework 1.0). Not only Internet Information Services (IIS) but also Personal web server Cassini (entered in the distributive compound) can act as the server of Web Data Administrator. For them, who are unfamiliar with it, let explain! Cassini - Web-server is with open source texts, written in C#. Cassini accepts the command only from local knots by default setup. For removing this limitation, need bringing some changes in the source texts of server, delete this fragment: // Limit to local requests only if (!_conn.IsLocal) {_conn.WriteErrorAndClose(403); return;} And do compile the server. While running Web Data Administrator, it opens a window shown in Picture-1 where one can choice Web server and Port, and will have access support. 9 4
3243BSSSN: Bit String Swapping Sorting Network for Reversible Logic Synthesis
Figure 1. Running window of Web Data Administrator After running Web Data Administrator one can open the home page of administrator in the browser by indicating appropriate address and port. After passing the authorization at the home page, an access to the own interface of administrator will appear. Here points out that interface it is simple and visible for users (administers of SQL server) so that they do not have to break their heads for any exercise (Picture-2).
Figure 2. Interface of SCDB of Web Data Administrator In the enlisted functions of Web Data Administrator present the controls of data base (Create and delete the data base, edit the tables structure and saving procedures, export and import data base from the users of SQL scripts, control the records of encountered users and groups (create, delete, edit). Except these Web Data Administrator includes SQL-client with the abilities to save the command texts in the file (This client is loaded when it crosses the foot-notes of query of objects in the tables of data base or in the control panel). This command result is reflected in the web page of SQL-client with the formatted form as shown in Figure-3.
III. ADVANTAGES OF WE DATA ADMINISTERING
The main advantage of using Web Data Administrator is its ability to remote administering of Microsoft SQL Server to those computers in which special client- program is not provided. Principally, the web interface of data-based server can be controlled with the help of Linux system. Wide possibilities of remote administering arise, of course, questions about security. But comfort and security are always contradictory criteria.
IV. THE PROGRAM CODE
At the end, lets see a little bit about programming. As mentioned above that Web Data Administrator is based on the technology .NET. The compound packet consists of two collections .NET (.NET Assemblies) - SqlAdmin.dll and SqlWebAdmin.dll. The collection of SqlAdmin.dll represents the main interest contain a series of class for accessing to the data-based server. The fragment of program is cited below (written in Delphi 8):
95
Md. Saiful Islam
Figure 3.The command result is in the table. Uses ..., SQLAdmin; ... procedure TWinForm.Button1_Click(sender: System.Object; e: System.EventArgs); var Server : SQLServer; begin Server := SQLServer.Create('Second'); Server.Username:='user1'; Server.Password := 'mushka'; Server.Connect; if Server.IsUserValid then begin RichTextBox1.Clear; RichTextBox1.AppendText( Server.Databases.Item['Northwind'].Script(SqlScriptType.Create)); end; Server.Disconnect; end; ... This program uses class SQL Server which represents the data-based server. This program is registered to the Server by indicating server name ("Second"), numbering the record of users and passwords, and then shows the output SQL-script, necessary for generating the data base of Northwind. Here field component is RichTextBox1. Work supplement is shown Picture-4. of Northwind. Here field component is RichTextBox1. Work supplement is shown Figure-4.
V. CONCLUSION
In this paper, we have discussed a mechanism for web-administering of databases. We proposed a tool Microsoft SQL Web Data Administrator to be used for this purpose which is helpful for SCBD. Cassini is a web server written in C#. It can act as web server where IIS are not present. We have also provided the necessary code written in Delphi 8 for web-administering using Web Data Administrator.
Figure 4. Example of clients supplement .NET for Microsoft SQL Server
VI. REFERENCES
[01] [02]
Zoltov . Global Servers. SPV.: BHV Saint-Peterburgh, 1998. 304 p. Semenov U. . Protocoles and Resources of SQL server. .: Internet Communications, 1996. 320 p.
96
[03] [04] [05]
Forolova . ., Forolov . . Web Server by our own hands. .: Dialogue-MIFI, 1997. (Library of Systems Programmer, . 29). Brown ., Zennett D. HTML 3.2 Original. SPV.: BHV Saint-Peterburgh, 1997. Florentine . ., Cherkovsky G. . Javascript. .: Research of Servers, 1998. (All Russian programmers, . 34). BHV Saint-Peterburgh, 1999. 720p.
[06] Tikhamirov Y.V. Microsoft MS SQL Server 7.0 as Original. SPV.:
Md. Saiful Islam1
ABSTRACT
In this paper, we have introduced the notion of UselessGate and ReverseOperation. We have also given an algorithm to implement a sorting network for reversible logic synthesis based on swapping bit strings. The network is constructed in terms of n*n Toffoli Gates read from left to right and it has shown that there will be no more gates than the number of swappings the 2 algorithm requires. The gate complexity of the network is (n ). The number of gates in the network can be further reduced by template reduction technique and removing UselessGate from the network. Keywords: Reversible Logic, Generalized Toffoli Gate, Hamming Distance, UselessGate, ReverseOperation and Template Reduction method.
I. INTRODUCTION
Power dissipation is a very important factor in VLSI design. The main benefit of reversible logic is theoretically zero power dissipation. A reversible circuit maps each input vector into a unique output vector. Landaurs principle [1] proved that logic computations that are not reversible necessarily dissipate heat irrespective of their implementation technologies. According to [2] zero energy dissipation would be possible only if the network consists of reversible gates. Thus reversibility will become an essential property in future circuit design.
Synthesis of reversible logic circuits differs significantly from the synthesis of combinational logic circuits. Because in a reversible circuit the number of inputs must be equal to the number of outputs, every output can be used only once (i.e., no fan-out is permitted), and must be acyclic.
has been proposed by Perkowski et al. [9] to realize symmetric functions. Sorting Network for reversible logic synthesis has been described in [10]. The idea here is used based on swapping bit strings. In fact one would expect that a better method could be found. In this paper, we have presented a convergent synthesis algorithm BSSSN and a variant of its Var_BSSSN based on swapping bit strings. The algorithm constructs a network of n*n Toffoli Gates read from left to right both for input and output translation. We have also given proofs why our algorithm will converge and introduced the notion of UselessGate and ReverseOperation that were not discussed previously. The paper is organized as follows: Section 2 provides some definition and basics necessary to understand our algorithm, Section 3 describes our algorithm, Section 4 explains how the network can be refined to reduce its circuit width, Section 5 will conclude the paper and references are cited in Section 6.
II. BACKGROUND
Although there exist many reversible gates in the literature good synthesis methods have not yet emerged. Miller et al have proposed transformation based algorithm with template reduction method in [5][6] and spectral techniques in [7] to find near optimal circuits. Mischenko and Perkowski in [8] have suggested a regular structure of reversible wave cascades and show that such a structure requires no more than the product terms in an ESOP realization of the function. A regular symmetric structure
An n-input n-output totally specified Boolean function (X), X = {x1, x2,, xn} is reversible iff it maps each input assignment to a unique output assignment. A reversible function can be written as a standard truth table as in Table 1 and can also be viewed as a bijective mapping 97
Md. Saiful Islam
of the set of integers 0,1,, 2n-1. Hence a reversible function can be defined as an ordered set of integers corresponding to the right side of the table, e.g. {1,0,3,2,5,7,4,6} for the function in Table 1. We can thus interpret the function over the integers as (0) = 1, (1) = 0, (2) = 3, etc. An n-input n-output gate is reversible if it realizes a reversible function. Many reversible gates have been proposed in the literature. One of the first gates was the CNOT gate [3], which capable of producing the exclusive or of two input bits as the second output and the first output is equal to the first input.
input-output patterns. For example, the value of C() for the function in Table 1 is 8.
Lemma 1. The upper and lower bound on the Hamming distance , between any two bit strings P and Q, in any reversible specification is n and 1, where n is the number of input lines. That is, 1 (P, Q) n. PROOF: LET P BE (A1,A2,,AN) AND Q BE SINCE IN A REVERSIBLE (B1,B2,,BN). SPECIFICATION NO TWO BIT STRINGS ARE IDENTICAL, THEY MUST DIFFER IN AT LEAST ONE POSITION. LET M BE THE INDEX AT WHICH AMBM. THAT IS, AM =BM, AND BM IS EITHER 0 OR 1. BIT STRINGS P AND Q MAY DIFFER AT MOST EVERY POSITION, I.E., AIBI, WHERE 1 I N. THUS WE CAN CONCLUDE THAT 1 (P, Q) N. LEMMA 2. TWO BIT STRINGS P AND Q CAN BE SWAPPED WITHOUT AFFECTING OTHERS IFF THE HAMMING DISTANCE BETWEEN THEM IS 1.
Table 1. 3*3 Reversible logic function A generalization of CNOT is a 3-input 3-output Toffoli gate [4]. The Toffoli gate negates the third bit iff the first two bits are 1. Figure 1 shows both gates as they are commonly drawn.
Proof: In any reversible specification bit string P and Q will be unique. Let P be (p1, p2, , pn) and Q be (q1, q2, , qn). Also let m be the index for which pm qm. That is, pi = qi, where 1 i n and i m. Since no bit string except P & Q will contain (p1, p2, , pm-1, x, pm+1, pn) where x may be either 0 or 1.If we use (p1, p2, , pm-1, pm+1, pn) as the control bits to drive the Toffoli gate, only P & Q will be affected. Example 2. In Figure 3, bit strings (1,1,1) and (1,1,0) have been swapped using T( b, c: a). Since the Hamming distance between them is one, this swapping can be carried out without affecting others.
Fig1.CNOTFig2.Generalized & Toffoli Gate Toffoli Gate
A generalized n*n Toffoli gate changes one bit, called the target, if some of the k bits are 1 which is shown in Figure 2. The changing bit, also called target, may also be in any position. The gate will be defined as follows T(xi1, xi2, ,xik : xn) where xn is the target and xi1, xi2, ,xik are the control bits. Garbage is the number of outputs added to make an n-input k-output function ((n, k) function) reversible. Given two bit strings, P and Q, the Hamming distance between them, denoted by (P, Q) is the number of positions for which P and Q differ.
Example 1. Consider the bit strings (1,0,1) and (0,1,1). The Hamming distance between these two bit strings is 2 since the number of positions for which these two bit strings differ is 2.
Fig 3. Swapping bit strings

UselessGate: A pair of gates in a network is considered to be useless if it has no effect in the circuit, i.e., when it is identified and removed from the network it will have no effect in the final output produced by the circuit. Example 3. T( c , b : a ) is an example of a useless gate in the gate sequence T( c , b : a ) T( c : a ) T( c , b : a ) and the gate sequence can be replaced by T ( c : a ) without any change or modification in the circuit. The final output
Given the function (X), the complexity C() is defined as the sum of the individual Hamming distances over the 2n 6
produced by the new circuit will be the same as the previous one.
Fig4:ExampleofUselessGate
for the function in table 1 will become {0,1,2,3,4,5,6,7} and the index of each element will be equal to itself. That is, if we define the set as {pi}, then 2n-1 i 0 and pi =i where n is the number of input lines. Section 3.1, 3.2, 3.3 and 3.4 will present such an algorithm named BSSSN (Bit String Swapping Sorting Network) that will construct a network as a sequence of Toffoli gates that might sort the elements of the set. The idea here is based on swapping bit strings whose hamming distance is exactly one as described in section 2.
B B P P B B
To identify and then to remove a UselessGate, we take into account the following property which follows directly from the definition of n*n Toffoli gates.
PROPERTY 2.1: A gate T( x1 , x2, , xk-1 : xk ) can be removed from the sequence T(x1 ,x2 , ,xk-1 : xk ) T( a1, a2,, al-1: al ) T(b1 ,b2 ,, bm-1 : bm) T(c1 , c2 , , cn-1 : cn) T(x1 , x2 ,, xk-1 : xk) iff xk {a1, a2, , al-1, b1, b2, , bm1,,c1 , c2 , , cn-1} and al, bm ,,cn {x1, x2,, xk-1}.
B B B B B B B B B B B B B B B B B B B B B B B B B B B B B B B B B B B B B B B B B B B B B B B B B B B B B B B B B B B B B B B B B B B B B B B B B B B B B B B B B B
1. Output translation BSSSN: Constructing the sorting network and its corresponding circuit
While (there is a bit string (a1,a2,,an) in the set that is not in its intended place, i.e., int_value(a1,a2,,an)index)
B B B B B B B B B B B B
ReverseOperation: A reverse operation places a bit string Z to its previous place if it is transferred to a different place by another operation(s).
Consider a sequence of operations Sl( Y, Z ) Sk( X, Z ) Sm( Z, Y ) in which Sm is the reverse operation of Sl, where operation S swaps its arguments.
B B B B B B B B B B
Example 4: In the example below let we want to swap (1,1,0) and (1,0,1). But we cannot do it directly since the Hamming distance between them is greater than one. This can be carried out by a sequence of swap operations: first (1,0,1) and (1,1,1), then (1,1,0) and (1,1,1). Finally (1,1,1) and (1,0,1).
Step 1. Let (a1,a2,,an) is not in its intended place then by induction its place is occupied by another string, say (b1,b2,, bn) Step2. Compute dist = ((a1,a2,,an), (b1,b2,,bn)) if dist=1, swap (a1,a2,,an) and (b1,b2,,bn) using gate T(a1,a2,, ak-1,ak+1, ,an : ak ) where akbk else{ find all the bit strings {(c1,c2, ,cn)}such that ((b1,b2,,bn) , (c1,c2, ,cn))=1 swap (b1,b2,,bn) with one of the(c1,c2, ,cn) such that ((a1,a2,,an), (c1,c2, ,cn)) will be minimum and in case of multiple (c1,c2, ,cn) low int_value bit string is chosen to break the tie and that is not in its intended place
B B B B B B B B B B B B B B B B B B B B B B B B B B B B B B B B B B B B B B B B B B B B B B B B B B B B B B B B B B B B B B B B B B B B B B B B B B B B B B B B B B B B B B B B B B B B B B B B B B B
if (c1,c2, ,cn) is in its intended place then do the ReverseOperation } Step3. Now (c1,c2, ,cn) will become the new (b1,b2,,bn) and goto step 2.
B B B B B B B B B B B B B B B B B B
Fig 5: Reverse Operation

III. SORTING NETWORK
Section 2 has described that a reversible function can be defined as an ordered set of integers corresponding to the right side of the table, e.g. {1,0,3,2,5,7,4,6} for the function in Table 1. Therefore, if we can build a network of reversible gates that might sort this set, it will eventually realize the function. For example, the ordered set of integers
Fig 6. Constructing network using BSSSN Figures 6 and 7 illustrate the application of BSSSN. The sequence of gates that form the network is: T(b,c:a) T(b,c:a)T(a,c:b)T(b,c:a) T(a,c:b).
Md. Saiful Islam
Fig 7. Circuit for the network constructed by BSSSN Algorithm BSSSN is straightforward. It is greedy in the sense it hopes that a bit string can be swapped by the one that is in its intended place. Because of Lemma 1 and 2, it is always possible to find two bit strings for Steps 2 that can be swapped. Therefore, it will always terminate successfully with a circuit for a given specification. The best case occurs when a bit string can always be swapped by the bit string placed in its intended position. A variant of BSSSN takes the bit string in the set whose integer representation is low and brings it to its intended place. Figures 8 and 9 illustrate the application of the variant of BSSSN and the sequence of gates that form the network is: T(b,c:a)T(b,c:a)T(b,c:a)T(a,c:b) T(b,c:a).
control lines to each Toffoli gate. For larger problems with up to 8 or 9 inputs this may not be a practical one. Selective use of control inputs can be used to swap elements. This can be carried out safely as long as it will not affect bit strings that are already in its intended place. We should choose a subset of the control inputs that will minimize the C() of the resulting specification. For example, we can select control inputs that will drive a Toffoli gate to bring more than one element at a time to their intended place.
Fig 10. Reverse Specification 4 Reduction Rules The circuits produced by BSSSN as described thus far frequently have gate sequences that can be reduced. For example, the sequence T(b:a) T(:b) T(:a) can be replaced by the sequence T(:b) T(b:a). Here we have implemented template driven reduction method introduced in [5]. In addition to template matching in [5] we have also removed useless gates that come in pairs and have no effect in the circuit.
IV. EXPERIMENTAL RESULTS
Fig 8. Constructing network using variant of BSSSN
Fig 9. Circuit for the network constructed by variant of BSSSN

2 Input Translation
For input translation, we have to find an inverse of the specification. For example, the reverse specification of the function in Table 1 is {1,0,3,2,6,4,7,5}. Then, we can apply BSSSN to realize the function. By applying BSSSN to realize the reverse specification we get the circuit: T(b,c:a)T(b,c:a)T(b,c :a)T(a,c:b)T(b, c:a) and its variant, we get: T(b,c:a)T(b, c:a)T(a,c:b) T(b,c:a)T(a,c:b)T(b,c:a).
3 Random Selection and Control input Reduction
In Section 3.1 we have shown some reversible examples and compare them with the circuits in [6][7]. Section 3.2 describes the method used to convert an irreversible specification to a reversible one and the way to synthesize them. Here we have synthesized a reversible circuit from an irreversible specification and not transformed an irreversible circuit to a reversible one.
1. Reversible examples
For each example, the specification is given as an ordered set of integers, which define the truth table specification of the reversible logic function to be realized. The circuit is given as an ordered sequence of Toffoli gates. Read from left to right they transform the left side to the right side.
To sort the elements in a specification, algorithms BSSSN and its variant take one element at a time and bring it to its intended place. Selection of a bit string randomly and then returning to its intended place can reduce the total number of swapping. This will minimize the total number of gates in the network also. Algorithm BSSSN also assigns the maximum number of 8
Example 3.1 Verification of realizing a Fredkin gate. This example is collected from [7]. The circuit given by our method produces the same result.
Specification: {0,1,2,3,4,6,5,7}
Circuit(BSSSN) : T(a:b) T(b,c:a)T(a:b) Circui(VAR_ BSSSN) : T(b:a)T(a,c:b)T(b:a)
i. ii.
Add n-k outputs each equal to one of the original inputs x1, x2,, xn-k. The k outputs will be realized on inputs xnk+1, xn-k+2, , xn.
B B B B B B B B B B B B B B
Example 3.2 This is a second example of the interchange of two positions in the specification. The circuit given by our method is identical to the solution provided by [8].
Specification: {0,1,2,4,3,5,6,7} Circuit(BSSSN) : T(a,b:c) T(a,c:b) T(b,c:a) T(a,c:b) T(a,b:c). Circui(VAR_ BSSSN): T(a,b:c)T(b,c:a)T(a,c:b) T(b, c:a) T(a,b:c).
It is easily verified that the specification constructed in this way maps an input pattern to a unique output pattern and is therefore reversible. The approach used by Miller et al. [7] adds unnecessary constant inputs and thus produces extra garbage outputs that are not actually needed to make the specification reversible.
Example 3.4 This procedure for transforming a singleoutput function is illustrated for the example of the 2-input EX-OR function in Table 2. The resulting circuit is the single gate T(a:b) which realizes the EX-OR of a and b on b. But the same function is realized in [7] with one constant on inputs and thus produces one extra garbage output that is unnecessary.
Example 3.3 This example is taken from [6]. The circuit given by our method is identical to the solution provided by the Bidirectional Algorithm in [6].
Specification: {7,0,1,2,3,4,5,6} Circui(VAR_ BSSSN): T(a,b:c)T(a:b)T(:a). 2. Non-reversible examples
According to [9] an irreversible function can be realized using reversible gates with the addition of some number of constant inputs and garbage outputs. The minimum number of garbage outputs required to transforming an irreversible function to a reversible one is log2m, where m is the maximum number of times a single output pattern appears in the specification.
B PB P
Table 2. (a) 2-input EX-OR (b) reversible specification derived from 2-input EX-OR using method described above (c) using method in [7]
Example 3.5 This example illustrates the realization of 2input AND function. The specification is {0,1,2,7,4,5, 6,3}. The solution produced by our algorithm is same to the solution provided by that of the [7], that is, T(a,b:c) which realizes the AND of a and b when c is 0 on input. Example 3.6 Full Adder Minimization: The resulting circuit produced by our algorithm is: T(a,b,d:c)T(a,b,d:c) T(a,b,d:c) T(a,b,c:d)T(a,b,d:c) T(a,b,c:d) T(a,b,c:d)T(a,b,c:d).
A single-output or a multi-output function involving input variables x1, x2, ,xn can be transformed to a reversible specification in the following way.
B B B B B B
a.
Compute m, where m is the maximum output pattern multiplicity of the irreversible specification. Let p = log2m and k is the number of outputs of . The value of k will be one for single-output function. b. If (p + k) > n Add (p +k-n) new input variable xn+1, xn+2, i. , xp+k and set each of them to zero on input in the circuit. Add n outputs each equal to one of the ii. original inputs x1, x2,, xn. k outputs will be realized on xp+1, xp+2, , iii. xp+k. iv. for single-output function replace by xn+1. Else
B PB P B B B B B B B B B B B B B B B B B B B B B B B B
which can be simplified by template matching to T(:d) T(a,b,d:c) T(a,d:c) T(a,b,d:c) T(b,d:c) T(:d) T(a,b,c:d)T(a, b:d)T(a,b,d:c)T(a,d:c)T(a,b ,c:d)T(a,c:d)T(a,b,d:c)T(b,d:c) (a,b,c:d)T(b,c:d)T(a,b,c:d). which can be again simplified by removing useless gates to T(:d)T(a,d:c) T(b,d:c)T(:d)T(a,b,c:d)T(a,b:d) T(a,b,d:c) T(a,d:c)T(a,b,c:d)T(a,c:d)T(a,b,d:c)T(b,d:c) T(b,c:d). After final simplification we get the following circuit: T(a,b:d)T(a:b)T(b,c:d)T(b:c). This circuit is identical to the circuit in [6][7].
Md. Saiful Islam
[10] Fig 11. Full Adder Though the initial circuit produced by our algorithm seems to be larger one, it can be simplified easily by simple template matching and identifying useless gates. Thus the circuit will be optimal. The main advantage of our algorithm is that it does not require exhaustive analysis like spectral used in [7].
V. CONCLUSIONS
[11] [12]
International Workshop on Logic Synthesis, 245-252, 2001. Md. Saiful Islam, Md. Rafiqul Islam, A. A. Mahmud and M.R. Karim, Sorting Network for Reversible Logic Synthesis, In Proc. of 3rd IEEE International Conference for Upcoming Engineers, ICUE 2004, Windsor University, Ontario, Canada, May 20-21, 2005. M. Perkowski, Private Communication. M. Neilson, and I. Chuang, Quantum Computation and Quantum Information, Cambridge University Press, 2000.
P P
In this paper we have given a convergent synthesis algorithm for reversible logic circuits. The specification must be totally specified. We have also given a method that will transform an irreversible specification to a reversible one. The algorithm always terminates with a network of Toffoli gates that can translate both input and output side to their corresponding output and input side. Since the synthesis of reversible circuits can be done in either side, this is valid.
VI. REFERENCES
[01]
[02]
R. Landauer, Irreversibility and heat generation in the computing process, IBM J. Research and Development, 5:183-191, 1961.
C.H. Bennet, Logical reversibility of Computation, IBM J. Research and Development, 17: 525-532, November 1973.
[03] [04] [05]
[06]
[07]
[08] [09]
R. Feynman, Quantum Mechanical Computers, Optic News, 11-20, 1985. T. Toffoli, Reversible Computing, Tech Memo MIT/LCS/TM-151, MIT Lab for Comp. Sci, 1980. D. M. Miller, D. Masolv, and G.W. Dueck, A Transformation Based Algorithm for Reversible Logic Synthesis, In Proceedings of the Design Automation Conference, pages 318-323, June 2000. D. Maslov, G. W. Dueck, and D.M. Miller, Fredkin /Toffoli templates for reversible logic synthesis, In International Conference on Computer Aided Design, November 2003. D. M. Miller, and G.W. Dueck, Spectral Techniques for Reversible Logic Synthesis, In 6th International Symposium on Representations and Methodology of Future Computing Technologies, pages 56-62, March 2003. Mishchenko and M. Perkowski, Logic Synthesis of Reversible Wave Cascades, In 11th IEEE/ACM Workshop on Logic and Synthesis, 2002. M. Perkowski, P. Kerntopf, A. Buller, M. Chrzanowska-Jeske, A. Mischenko, X. Song, A. Rabadi, L. Joswiak, A. Coppola, and B. Massey, Regularity and Symmetry as a base for efficient Realization of Reversible Logic Circuits, In 10th
P P P P
10
CALL FOR PAPERS

The Faculty of Computer Science is on the verge of publishing the Volume 1, Number 2,
the afore mentioned journal for the December 2007 issue.

SUBMISSION DATE: Last Date of submission Notification of Acceptance Last Date of Final submission 30th August 2007 15th October 2007 30th October 2007
P P P P P P
December 2007 issue of the Journal of Computer Science. Each issue of the Journal of Computer Science of IBAIS University is devoted to all the recent technical topics in the field of computer science and engineering and thus provides to Journal of Computer Science readers a
SUBMISSION GUIDELINES:
The following points may be treated as broad guidelines for submission of the articles/paper:
collection of up-to-date technical papers, book reviews, research papers and innovative finding on those topics. These issues are valuable to the computer science research community and
1. 2.
The technical papers should be preferably research-based and should demonstrate the authors own thought and judgment. The manuscript must be both in hardcopy and softcopy ( MS-Word), single spaced on one side of A4 size paper with generous margin, and should not exceed 2,500 words ( including footnotes) or 15 pages in length and be sent in duplicate together with a softcopy submission through *.doc ( MS-Word ) . There should be an Abstract given at the top of the text. The text should be organized under appropriate headings/titles and subheadings/subtitles. Tables should be given separate numbers in the manuscripts, such as Table 1, Table 2 etc. Figures should be given separate numbers in the manuscripts, such as Figure 1, Figure 2 etc. Illustrations, graphs and charts must be computer print out. No drawing papers are accepted. Name(s), and designation of the author(s) with the title of the technical papers should be provided on a separate sheet
become valuable references. The technical topics such as artificial adhoc intelligence, networking, teletraffic artificial
engineering,
intelligence, robotics, pattern recognition, image processing, MIS, database management system, computer networks, theory of computing, metrics,
4. 5. 6. 7. 3.
telecommunication,
software
experimental and theoretical aspects of material research, electronic, communication, computer science and engineering would be covered by Journal of Computer Science of IBAIS University. Journal of Computer Science of IBAIS University publishes only papers that are submitted in response to a Call-for-Papers.
All the interested authors teachers, academicians, researchers and others belonging to the university level educational institutes and research organizations have been requested to submit articles/papers for publication in
Bibliographical references should be made in the following defined structures: 1. In the case of Articles, the citation should be as follows:
11
Md. Saiful Islam
Authors Last name, First Name and Middle Initial, Year of Publication. Title of the Article in double inverted comma, Name of the Journal (to be italicized ) Volume Number ( Month ). Page Number. 2. In the case of Text Books, the citation should be as follows: a. Authors Last name, First Name and Middle Initial, Year of Publication. Title of the Book (to be italicized ) . place, name of the Publisher and Page numbers
a.
SUBMIT TO:
The submitted manuscript has not been published elsewhere, and is not being considered for publication elsewhere in any form. Manuscripts are reviewed and selected by TWO anonymous reviewers as determined by the Board of Editors. The authors are requested to submit two (2) copies of hardcopy (computer paper printout ) and one (1) copy of softcopy ( *.doc) by postal service, or by drop off or by electronic submission via e-mail to:
Professor Md. Matiur Rahaman Mian, PhD Chief Editor Board of Editors Journal of Computer Science IBAIS University House No 21/A, Road no 16 (Old 27) Dhanmondi, Dhaka 1209 e-mail: jcs@ibais.edu
or, Sohail Ahmad Technical Editor Board of Editors Journal of Computer Science IBAIS University House No 21/A, Road no 16 (Old 27) Dhanmondi, Dhaka 1209 e-mail: sohail@ibais.edu
12

Volume 1 No 1

Transféré par

Informations du document

Titre original

Copyright

Formats disponibles

Partager ce document

Partager ou intégrer le document

Options de partage

Avez-vous trouvé ce document utile ?

Ce contenu est-il inapproprié ?

Droits d'auteur :

Formats disponibles

Volume 1 No 1

Transféré par

Droits d'auteur :

Formats disponibles

2007 Journal of Computer Science Vol 1, No 1, June 2007, pp 05-10 Faculty of Computer Science ISSN 1994-6244

Capacity and Coverage Calculation Model for the UMTS

Md. Amiruzzaman et al.

Md. Amiruzzaman et al.

given cell [9]. That is to say,

(9) Where A, B are constants defined by the frequency and

Uplink 173.93dBm/Hz 3dB 170.93dBm/Hz

Md. Amiruzzaman et al.

Table1: Typical link budget for UMTS cell

Md. Amiruzzaman et al.

Md. Saiful Islam et al.

Figure 2(a): Client-Server Connection Protocol

Md. Saiful Islam et al.

and there are N labeled training

( x1, y1), ( x 2, y 2),..., ( xN , yN ), x R d w

Department of Mathematics, University of Dhaka, Dhaka-1000. Email: msa417@hotmail.com.

Md. Showkat Ali.

subject to the constraint

support points of the two classes. We can equivalently minimize

A Note On Symplectic And Contact Geometry

Md. Showkat Ali.

Theorem: Given a linearly separable training examples

A Note On Symplectic And Contact Geometry

Md. Showkat Ali.

A Note on Symplectic and Contact Geometry

: V V . This means that the following conditions are satisfied:

Skew-symmetry: for all

for u , v V . The linear symplectomorphisms of (V , )

( F + G) = F G , (iv) ( F G ) = F + G , (v) dim F = dim V dim F .

where a, b = 1,2, L , n and such a basic { X , Yb } is called symplectic frame.

2 ( Z a , Yb ) = a . Put ab = ( Z a , Z b ) and X a = Z a + ab Yb (with the summation convention).

Matiur Rahaman Mian

V and (V , ) is a symplectic vector space.

For the proof of this Proposition, let

denote the projection along

v x = 0. x1 , x 2 , ... , x n , y1 , y 2 , ... , y n equipped with a differential

If we consider the space M = R2n with coordinates form =

dy i . Then the pair ( M , ) is a symplectic manifold.

(U , x1 , x 2 , ... , x n , y 1 , y 2 , ... , y n ) centered at p such that on U,

(U , x1 , x 2 , ... , x n , y 1 , y 2 , ... , y n ) is called a Darboux chart.

choice of local trivialization; for if

is any 1-form, d ( f ) = f d + df . However, it is now clear that

d D thought of in this way, is exactly

2n + 1 3 . Moreover, the non-degeneracy condition exactly requires that [6]

d D is to be non-degenerate, D must have positive even rank

x = ( x1 , x 2 , ... , x n ), y = ( y1 , y 2 , ... , y n ) and z- the local coordinates.

Matiur Rahaman Mian

{U i } be an open covering of M. We call M a complex

U i U j is nonempty, then there exists a nonvanishing holomorphic function f ij on U i U j such that i = f ij j on U i U j .

i = f ij j , it follows easily that

i (d i ) n is a holomorphic form of degree 2n + 1, { f ij ( n +1) } defines the canonical bundle K of M. The

characteristic class of K is c1 ( M ) . The line bundle E = T ( M ) / F is defined by obtain

= (n + 1) , it follows that c1 ( F ) = n . Hence completes the proof.

{ f ij 1} and p the projection of L onto M. Let

hi : p 1 (U i ) U i C be the coordinate map. If v L x and x U i U j , then

showing that { z i . p ( i )} defines a holomorphic 1-form

on L. A simple calculation shows that

A = (2n + 1)! /(n!) 2 .

i (d i ) n i (d i ) n db is a form of degree 4n + 3 (> real dimension of M) on M, it vanishes identically.

Hence, the restriction of

2 Ab 2n z i dz i p * ( i (d i ) n i (d i ) n ), which is clearly different from zero at every point P.

z 1 , z 2 , ... , z 2 n +1 , z 2 n + 2 be a coordinate in the (2n + 2)- dimensional

= z1dz 2 z 2 dz1 + L + z 2n+1dz 2n+ 2