NATIONAL INSTITUTE OF

TECHNOLOGY,ROURKELA

A SOFT COMPUTING PROJECT

ON

USER AUTHENTICATION USING NEURAL

NETWORKS
GUIDED BY:-
Prof. A.K SAHOO

SUBMITTED BY:-
BHAGABAT PRASAD SAHOO
ROLL-217EC5325
M.Tech(C & N)
 CONTENTS

1. INTRODUCTION

2. THEORY

3. EXPERIMENT

4. MATLAB CODE

5. OUTPUT

6. OBSERVATIONS

7. CONCLUSION

8. REFERENCES
 INTRODUCTION
USER ID AUTHENTICATION AND NETWORK SECURITY

User authentication is very important in a networked smart environment. Authentication

schemes are mostly based on passwords, smart cards and biometrics. Network security is a
concept to protect data transmission over a wired or wireless network. The easiest approach
to safeguard a network device from unauthorized access is to assign him a different user id
with a matching password. The usage of the technology and advancement in the technology
is directly related to the ascending graph of the wireless networking. However, side by side
intruders and unauthorized persons will also try to breach the security measures. This leads
an increasing need for taking extra security measures in wireless technology.

Network security considers authorization of access of data in a network. Password

authentication is one of them. Every user assigns an Id and corresponding password
that allows them to access information without any problem. This Id is the identity of its
authorized usage

NEURAL NETWORKS
A Neural Network is consist of set of parallel and distributed processing units called
neurons. These neurons are connected by unidirectional or bidirectional links in an ordered
fashion i.e.in layers. Three kinds of layers are present in Neural Network architecture, input
layer consists of n number of inputs to the network and each input is multiply by a weight
associated with this input. The product of input and weight are summed and feed into a
network through a transfer function to generate the output.
Neural Network designing is based upon network topology, network transfer functions and
network learning algorithms.

a) Network Topology:
e.g. number of hidden layers, number of neurons in each hidden layer, types of activation
function, concept of over fitting and under fitting, interconnection within the layer.
Two important neural network topologies are Feed Forward topology and Recurrent
topology.

b) Network Transfer Function:

The transfer functions and activation functions play an important role in learning algorithms
and influence the complexity and performance of neural network. There are various transfer
functions such as linear, sigmoid, polynomial etc.

c) Network Learning Algorithm:

There are three types of network learning algorithms: supervised, unsupervised and
Reinforcement.
 THEORY
Neural Networks are used as an encryption and decryption in cryptography .Topology of
each neural network is based on their training sets. Generally encryption method is done on
the input data and the decryption method is done on the output string of digits. If we use
neural network for the network security means for encryption and decryption, their keys
have adapted neural network parameters: their architecture and their configurations.

The proposed authentication scheme includes login and authentication procedures. The user
chooses the user Id (ID) and password (PW). These numbers are used to train the neural
network. The neural network repeats the process again and again (iterations) until the
network can remember the entire user Id and its corresponding password pairs. This
password authentication scheme is based on Back Propagation Neural Network and it
describes the verification of user Ids and passwords.

BACK PROPAGATION ALGORITHM:

Backpropagation, short for "backward propagation of errors," is an algorithm for
supervised learning of artificial neural networks using gradient descent. Given an artificial
neural network and an error function, the method calculates the gradient of the error
function with respect to the neural network's weights
 EXPERIMENT

TOOLS:- MATLAB R2017b

PROCEDURE:
Here we use a predefined set of user id and password combinations and prepare a neural
network for 8 users.

USER ID PASSWORD
[0 0 0] [1 1 1]
[0 0 1] [1 1 0]
[0 1 0] [1 0 1]
[0 1 1] [1 0 0]
[1 0 0] [0 1 1]
[1 0 1] [0 1 0]
[1 1 0] [0 0 1]
[1 1 1] [0 0 0]

We then use MATLAB NNTOOLS(Neural network tools) to create a Back progation

algorithm network that adapts to the user id and password combinations.

1) USER ID AND PASSWORD DATA INPUT:

2) NEURAL NETWORK CREATION:

Network type:- Feed forward back prop

No. of neurons:- 4
Activation function:- Tan sigmoid
3) NETWORK TOPOLOGY:

4) TRAIN THE NETWORK:

5) SIMULATE THE NETWORK:

6 ) ADAPT THE NETWORK:

7) GET THE WEIGHT VALUES:

MATLAB CODE
 We derive the values of the weight and bias of each layer from the above trained
neural network.
 Here we use these values to determine the corresponding password output for each
user id entered by the user.
 Then the entered password is matched with actual password generated by the
network.
 If both are same then ‘verified’ is generated and if they are different then
‘unauthorized’ is generated.
OUTPUT

Let the user id be [100] with the corresponding password [011].

So for the following combinations we get the corresponding outputs.

OBSERVATION
We observe that the program gives ‘unauthorized ’ as output for all combinations except for
the correct one i.e [0 1 1] for which it gives the output ‘verified’.

CONCLUSION
From the above observation we can conclude that the MATLAB program indeed
successfully authenticates User id and password using back propagation algorithm.
REFERENCES

1) Dahiya, Menal & Gill, Sumeet. (2018). Securing Password For Wireless Device Using
Training Functions of MATLAB. International Journal of Pure and Applied Mathematics.
118. 9-19.

2) Reyhani, Shahbaz & Mahdavi, Mehregan. (2007). User Authentication Using Neural
Network in Smart Home Networks. International Journal of Smart Home.