Vous êtes sur la page 1sur 6

CROSS - PLATFORM SOCIAL MEDIA NETWORK FIND ANONYMOUS USER

Ms. Samundeeswari K, Mrs. Geetha S.,

M.phil Research Scholar, Department of computer Science, Muthurangam Government Arts College
(Autonomous), Vellore, TamilNadu, India.
Associate Professor, Department of Computer Science Muthurangam Government Arts College (Autonomous),
Vellore.
---------------------------------------------------------------------***---------------------------------------------------------------------
Abstract vast tests illustrate that FRUI is better than the methods
based on the cut-term network architecture.
Over the past two years the social media
network (SMN) has seen a vibrant research stream Keywords: UMP-User Matched Pairs, FRUI- Friends
expression and evolution in a larger style of computers. Relationship based User Identification, Social Media
There is a problem that cannot identify some of the Network (SMN),
SMNs, more and more identical customers. A clear
Introduction
study of social networks helps to solve many issues and
problems in packages. Because the general aims can be Over the past decade, several types of social networking
duplicated by customers for different purposes and can sites have emerged and have contributed greatly to the
be easily implemented, textual editing of users' public real world ranking of social behavior. Twitter 1, the
profiles are particularly weak in most modern person- largest micro blog service, has over 600 million users
recognizing resolutions in the mine. Some research is and produces 340 million tweets per day [1]. The Sina
aimed at users based on consumer content and writing Micro blog 2, the primary Twitter-style Chinese Micro-
styles. However, most of the SMNs have places where blog website, has over 500 million accounts and
there are places, and the style of writing from the quick produces more than 100 million tweets per day [2]. Due
penalties of leading SMNs, including Cine to this diversity of online social media networks
Microbiology and Twitter, is difficult. Furthermore, (SMNs), they use different SMNs for different
since the on-line SMNs are pretty symmetrical, it is not purposes. For example, RenRen 3, Facebook style but
entirely beneficial to fully-based consumer antonymous SMN, used in China for Blogs, is used to
identification programs in network form. The real world share Sina Micro blog positions (Figure 1). In other
scars are often compatible milk cycle in the characters words, each resident SMN meets some user needs.
and the percentage of the consumer clearly. Therefore, it Based on SMN management, you can provide
is perfectly right to use a friendly structure to study integrated information for each user that applies to
urban-type SMNs. Since Equal Customers set up anonymous users in various SMN sites, and can adapt to
partially similar friendly systems on the extraordinary relevant terms such as the terms of service. In theory,
SMNs, we put forward the Relationship User Identity cross-stage research allows a bird eye view of SMN
(FRUI) algorithm. FRUI calculates a contest diploma user behavior. However, almost all recent SMN-based
for all candidate user-guided Pairs (UMPs), and the studies focus on an SMN site, providing incomplete
simplest UMPs with the highest standard are considered data. Therefore, the study examines the strategy of
equivalent to customers. We have created two proposals crossing several SMN sites to draw a detailed picture of
for improving algorithm efficiency. The results of the these behaviors.
Nevertheless, cross-platform research meets commentary tunnel attempts to analyze communities
many challenges. As shown in Figure 1 with the through various technologies (Fluke, Lawrence, & Giles
development of SMN sites on the Internet, the cross- 2000, Hu & Li 2004). However, in most cases, the
platform approach integrates various SMN sites, analysis is limited to a community. Cross-community is
producing rich source data and SMNs for social a big problem when dealing with any kind of analysis of
computing tasks. SMN users set up natural bridges on these communities is disconnected. The disappearing
these SMN sites. Cross-platform The main heading for element is a link to the users in different communities,
SMN research is user identification for various SMNs. which is a key factor in any link analytics algorithm.
The research of this topic sets the foundation for cross- This is because of the inability of the Internet and most
platform SMN research. User identity is also called user communities to protect users' name by selecting users
authentication, user identification resolution, user fit instead of using different usernames and authentication
and anchor linking. Although any solution can not systems instead of their actual identities. In addition,
identify anonymous SMN users, some SMN communities rarely share single-login practices, and
components can be used to identify a section of users in users can login to different communities using one
many SMNs. Several studies, [8], [9], [9], [9], [9], [9], username (e.g., Orkut and YouTube). However, if there
[7], [6], [6], [7] [10], [11], [12], [13], [14], [15], [16], is a mapping between different communities and the
[17]. These attributes do not require the uniqueness and real identities behind them, it is a straightforward task to
these programs are very weak because of the various integrate communities across the web. Can we see this
purposes (including malicious users) easily deceptive to map? In this paper, we provide evidence on the
users. Some researchers need to recognize users using existence of this mapping, and prove that we are a step-
the poster, location and writing style [18], [19], [20], by-step process to identify all relevant identities across
[21], [21], [21]. These techniques are affected by communities. We present the Identity Identity Issue for
limitations because the location data is difficult and the first time, and then forward the experience to the
difficult to separate from narrow sentences because of user's behavior on the Internet, discuss the proposed
the style of writing. While almost all SMNs are method for identifying the next identifying marks, and
collected and difficult to attach, we only identify user then refer to our test results and results.
users [22], [23], [24] if we review our literature review.
Data security and privacy, anonymized a social net-job
map with links to known identities [24] based on
Narayanan and Shmatikov (abbreviated NS). NS is the
first attempt to fully authorize users by using the
messages, and successfully applies to 30% accounts
Connecting Users across Social Media Sites
with a 12% error rate. Parthenov and many others. [23]
Someone verifies adult when someone is trying
The joint-attachment attribute was proposed.
to determine whether "a 11-year-old girl or a 45-year-
Literature Survey old man." The New York Times [16] pointed to its
Data Mining and Social Media Analysis
significance, saying "Scot, a mobile social networking
(Wassman & Fast 1994) Social analysis is an interesting
apparatus found that three adults aged between 13 and
issue in recent developments. Here, a community refers
17 were within two weeks, in three separate incidents,
to a particular social media website (eg, good). It is
they contacted children, police said and sexually
noteworthy that the current analysis, link analysis and
assaulted." Age verification Judgment is a crucial issue.
In 2008, an Internet Security Technology Program was While both the screen name and profile image can be
initiated to evaluate aging verification technologies identified by users, they can not be used for larger
when meeting with academics and web companies. The SMNs. This is because some users have name and
same report states that "four years later, the members of profile pictures on the same screen. For example, many
the console were very well oppressed," and "an users have the screen name "John Smith" on Facebook
informal analysis of key statistics in the field of Content-Based User Identification
artificial intelligence is no research". The New York
Content-based user identification solutions will
Times reports that age verification is more difficult for
attempt to identify users based on the method and
social media, where people who do not fit into two
location of content, as well as the writing style of the
social networks can expect an unidentified algorithm
content.
(JLA) and get a portion of the same users. Gorla and
Network Structure-Based User Identification
Latani [24] used the number of unmapped users degrees
and the number of common neighbor countries to fix Network-based analysis based on user identity
SMNs. from multiple SMNs is used by single user users to

Related work identify the same users. These are user network

Profiles, contents and network structures are configurations and seed or anchor, identified users. As

three cardinal components in an SMN. Accordingly, shown above, a network-based user's identity presents

current studies on cross-platform user identification can several important challenges; some studies have to be

be divided into three categories: profile-based, content- created.

based and network structure-based approaches. Methodology

Profile-Based User Identification  Cross-Platform In SMN’s


 Anonymous Identical User
Many of the anonymous user identifiers focus
 Friends And Relation
on public profile attributes that include screen name,
gender, birthday, city and profile picture. A general
Algorithm 1: FRUI
name common to almost all SMNs is a screen name.
The way to recognize users of various SMNs was Input: SMNA, SMNB, Priori UMPS: PUMPS
widely reviewed. Perito and many others [3] calculate
Output: Identified UMPS: UMPS
the screen names and the identities of identified users
1: function FRUI (SMNA, SMNB, PUMPS)
using binary typewriters. Similarly, Liu et al. [4] Using
2: T = {}, R = dict(), S = PUMPS, L = [], max = 0,
screenshots he applied users to an unsupported
FA = [], FB = []
approach. Japarani and Liu [5] have been proposed to
3: while S is not empty do
identify multiple SMN platforms, and have been
4: Add S to T
asserting many assumptions. Upon this job, they created
5: if max > 0 do
[6] and the user's behavior on the user's screen by
6: Removes S from L[max]
changing the user's mapping system. In the public
7: while l[max] is empty
profile attributes, the profile picture is another aspect of
8: max = max – 1
considerable research. Ed. [7] Identified the task of
9: if max = = 0 do
identifying user with the face identification mechanism.
10: return UMPS
11: Remove UMPS with mapped UE from 2: while VerterList ≠ 0 do
L[max] 3: let SeedVerter = VertexList.head() and remove it
12: foreach UMPA~B(i,j) in S do fromVertexList;
13: foreach UEAa in the unmapped neighbours 4: for each vi € VertexList do
of UEAi do 5: calculate Cost (SeedVertex, v i ) using the
14: FA [I] = FA [I] + 1 anonymization method for two variables;
15: foreach UEAb in the unmapped 6: end for
neighbours of UEAj do 7: if VertexList.size() ≥ 2K – 1 then
16: R[UMPA~B(a,b) + = 1, FB [J] = FB[J] + 8: let CandidateSet contain the top k – 1 vertices
1 with the smallest Cost;
17: Add UMPA~B(a,b) to 9: else
L[R[UMPA~B(a,b)]] 10: let CandidateSet contain the remaiming
18: if R[UMPA~B(a,b)] > max do unanonymized vertices;
19: max = R[UMPA~B(a,b)] 11: end if
20: m = max, S = {} 12: suppose CandidateSet = {u 1,........,um}, anonymize
21: while S is empty do Neighbor(SeedVertex) and Neighbor(u1);
22: Remove UMPS with mapped UE from 13: for j = 2 to m do
L[max] 14: anonymize Neighbor(uj) and
23: C = L[m], m = m – 1, n = 0 {Neighbor(SeedVertex), Neighbor(u1),.....,Nerghbor(uj -
24: S = {un-Controversaial UMPS in C} 1)}, mark them as “anonymized”;
25: while S is empty do 15: update VertexList;
26: n = n + 1, I = {UMPS with top n Mij in 16: end for
C using (5)} 17: end while
27: S = {un-Controversaial UMPS in I}
28: if I = = C do System Architecture
29: break
30: return T

Algorithm 2: Anonymizing a Social Network to


achieve k-anonymity.

Input: a Social network G=(V,E), the anonymization


requirement parameter k,
The cost function parameters α,β and γ;
Output: an anonymized graph G`;
Initialization: initialize G` = G, and mark vi € V(G) as
“ unanonymized”;
1: sort vi € V(G) as VertexList in neighbourhood size
descending order;
Fig: System Architecture importance and helps to resolve the de-imaging user
identity tasks. So, we proposed a consistent network
Anonymizing a Social Network
system based user identification solution. The novel
First, we refer to all the Verdes in the network friend of the so-called FRUI developed a relationship-
"without knowledge". We have a page in a layer of based algorithm. To improve the efficiency of the FRUI,
"unrecognized" vertical look. In the order of the queue we wrote two proposals and told us the complexity.
size line: vertices u, v ∈ V (G), v if (V. <| V) |, or | V Finally, we check our algorithms for artificial artificial
(eight ant (u)) | = | V (A Embark G (V)) | And | (A webs and ground-based networks. The results of our
Embark G (U)) | <| If E (N Emperor G (v)), then their experience tests reveal that the net job structure can
surroundings and the same size angles and ends before fulfill the important user id work. Our FRUI algorithm
they are listed, they can be ordered independently. is simple, more efficient and better than NS, the existing
state art network configuration user solution. When the
Obviously, we first select SeedVertex on the
raw text data is less than complete, incomplete or hard
VedexList list. The cost of animation of SeedVertex and
to be obtained by the privacy settings, the FRUI is more
any other verteks is calculated using the two vertical
suitable for cross-functioning tasks. The results of our
anonymous method discussed in Section 4.2.2. If the
experience tests reveal that the net job structure can
number of unnamed verticals in the VertexList is at least
fulfill the important user id work. Our FRUI algorithm
2k - 1, we choose a set of candidate sets for the first k-1
is simple, more efficient and better than NS, the existing
vertex in VertexList at a very small cost. We can pay
state art network configuration user solution. In
less in the amount of animations based on the number of
situations where the source text data is scattered,
rounds and the number of ends. Since all the cuts in
incomplete, or hard, the FRUI crosses the cross-tasking
VertexList are smaller or equal to SeedVertex than the
tasks.
SeedVertex, we scan VertexList in the layer of the sheet
size and the lowest amount of anonymous expenditure References
is the current (k - 1) - a similar reward.
[1] Andrey Bogomolov, Bruno Lepri, Michela
Seventarex and candidates candidates = {u1,. . , Ferron, Fabio Pianesi,and Alex Pentland. Daily stress
um} is anonymized using the two-triangulation recognition from mobile phone data, weather conditions
anagation method in the partition 4.2. The anonymity of and individual traits. In ACM InternationalConference
SeedVertex and u1 is straight forward. After these two on Multimedia, pages 477–486, 2014.
patches anonymous, their surroundings are the same. As
[2] Chris Buckley and EllenM Voorhees. Retrieval
for u2, any change near SeedVertex (in order to add one
evaluation with incomplete information. In
edge or a neighbor's edge) may be used for U1, which
Proceedings of the 27th annual international ACM
means that SeedVertex, u1 and u2 both are same. This
SIGIR conference on Research and development in
process continues until SeedVertex and u1,
information retrieval, pages 25–32, 2004.
surroundings. . . , um are anonymous.
[3] Xiaojun Chang, Yi Yang,Alexander G Hauptmann,
Conclusion and future work
Eric P Xing,and Yao Liang Yu.
This review provided an innovative solution to the user
Semantic concept discovery for large scale zero shot
identification problem on SMN sites. As an important
event detection. In Proceedings of International
aspect of SMN, the network system is of utmost
JointConference on Artificial Intelligence, pages 2234-
2240, 2015.

[4] Wanxiang Che, Zhenghua Li, and Ting Liu.


Ltp: A Chinese language technology platform. In
Proceedings of International Conference on
Computational Linguistics, pages 13 16, 2010.

[5] Chih chung Chang and Chih-Jen Lin. Libsvm:


a library for support vector machines. ACM
TRANSACTIONS ON INTELLIGENTSYSTEMS
AND TECHNOLOGY, 2(3):389-396, 2001.

[6] Dan C Ciresan, Ueli Meier, Jonathan Masci,


Luca Maria Gambardella, and J ̈ urgen
Schmidhuber. Flexible, high performance
convolutional neural networks for image classification.
In Proceedings of International Joint Conference on
Artificial Intelligence, pages 1237- 1242, 2011.

[7] Sheldon Cohen and Thomas A. W. Stress, social


support, and the buffering hypothesis. Psychological
Bulletin , 98(2):310 –357, 1985.

[8] Glen Coppersmith, Craig Harman, and Mark


Dredze. Measuring post-traumatic stress disorder in
twitter. In Proceedings of the International Conference
on Weblogs and Social Media, pages 579–582, 2014.

Vous aimerez peut-être aussi