MK-means clustering applied to user authentication

In this research work, the authors have been proposed MK-means clustering algorithm to enhance the security features of biometric user authentication system with continual learning. The proposed algorithm enhances user authentication by deterring the password sharing. It is used to train individual, per-keyboard profiles that is based on keystroke frequency of users while typing on computer keypad. The important factor here is the system learning is continuous, more and more use the system the security increasing. These profile use Data Encryption Standard (DES) encryption algorithm, read at log-on time of actual password key by using the access control techniques to further verify the identity of a user. MK-means values are obtained from the membership functions applied to the keystroke frequency that have been compared against the profile values and a match within a certain precision threshold. If both the parameters are matched, the user grant the access, otherwise, not. The proposed algorithm efficiently searches the number of clusters and their respective location, to optimize the measure. The experiments proved that MK-means provides more security as compare to K-means clustering algorithm.