*WINNER* Machine Learning with Feature Selection for Continuous Authentication

Abstract

This won best graduate poster in Computer Science.

Single point of authentication can serve as security weakness for highly secure systems. Although there are advanced authentication methods such as graphical passwords, one time passwords and bio metric authentication, they still face the same problem of single time authentication where masqueraders can potentially hijack sessions as authorized users. To address this problem, continuous authentication monitors user’s activities right from the time the user logs into the system until the user logs out. Machine learning has been used for continuous authentication that has the ability to learn useful knowledge about authorized user without direct programming. Rather than manually selecting features to feed machine learning algorithms, feature selection algorithms can be used to yield better results. In this project, we have experimented with various combinations of feature selection and machine learning algorithms to learn models of authorized user keystroke dynamics during the authentication process. Preliminary results show that such additional level of data analytics helps to improve unauthorized user detection.