Authentication is the process of identifying
an individual, usually based on username and password. Authentication merely ensures
that the individual is who he or she claims to be. This forestalls the
activities against confidentiality and integrity. The traditional and common
approach for authentication method is to use alphanumerical passwords. It is
known to have some Drawbacks. Mainly Users tend to pick passwords that can be
easily remember [3]. On the other hand, if a password is hard
to guess, then it is often hard to remember. To address this problem, some
researchers have developed authentication methods that use Graphical passwords
[9]. These graphical passwords [7, 9, 10, 11, 13] are also suffered due to shoulder-surfing
and screen-dump attacks [9]. For different scenarios we use different
authentication techniques.
Shoulder surfing is the main
problem of graphical passwords [7, 9, 10, 11, 13]. To overcome the problem of
shoulder surfing we introduce a novel scheme called S3PA [1]. This S3PA scheme
provides the login screen to the user at every time the user logs in, this
login image consists of set of characters. User with his password clicks some
pass characters which are different for different sessions and explained in
proposed scheme. So we are providing a mechanism to overcome shoulder surfing.
The main limitation in using the
traditional password authentication method is that, a server must maintain a password
table that stores each user’s ID and password [24]. If an intruder attacks the
server system through attacks like SQL injection then he can access the entire
information from the table. Even if the information is stored in the hashed
format [24] or encrypted format the intruder can affect the system like appending
the wrong password for an identified user etc. hence to overcome this setback
we use soft computing approach, a neural network with Back Propagation Learning
[2].
In our scheme users are allowed
to give a username and password at the registration phase without any password
policies, we are going to develop a Feed Forward Neural Network which takes the
user name as an input and produces our password which was given at registration
as output [2]. Hence we have to perform training to the network. There are
different training algorithms for training a neural network , we are going to
employ a supervised learning algorithm which is Back propagation [5]. The
weights of the neural network are stored at the sever side DB [2,4]. Hence
every time the user logs in he provides his/her username, this username is sent to server, and
then server gives the username as input to the FFNN with the stored weights as weights of the
neural network therefore providing the password of the user. This obtained
password is synchronized with our S3PA scheme [1]. An image is sent to the user
which contains complete set randomly scattered characters. Then user finds his
pass characters through our schemes and enters the every pass character formed
by pass triangles, the coordinates of these are sent to server then server
checks the authenticity of the user.
In this thesis we are going to
explain different password mechanisms and different stages of password
enhancements, drawbacks of each mechanism. Then we are going to discuss the
concept of shoulder surfing, how we introducing our scheme to overcome shoulder
surfing. We will provide a detailed
explanation of S3PA scheme. Then followed by Concepts of neural network, like
defining and developing a neural network, training the neural network and how
we provide liaison between this and S3PA scheme.
No comments:
Post a Comment