A Probabilistic Approach for Authenticating Text or Graphical Passwords Using Back Propagation

Password authentication is a common approach to the system security and it is also a very important procedure to gain access to user resources. In the conventional password authentication methods a server has to authenticate the legitimate user. In our proposed method users can freely choose their passwords from a defined character set or they can use a graphical image as password and that input will be normalized. Neural networks have been used recently for password authentication in order to overcome pitfall of traditional password authentication methods. In this paper we proposed a method for password authentication using alphanumeric password and graphical password. We used Back Propagation algorithm for both alphanumeric (Text) and graphical password by which the level of security can be enhanced. This paper along with test results show that converting user password in to Probabilistic values enhances the security of the system

💡 Research Summary

The paper proposes a password‑authentication scheme that replaces the traditional password table with a feed‑forward neural network trained by the Back‑Propagation (BP) algorithm. Users may choose either an alphanumeric (text) password or a graphical password (an image). In both cases the chosen password is first converted into a numeric representation: for text passwords each character is assigned a unique integer from a defined character set (e.g., ASCII) and then normalized to a probabilistic value in the range