EXPLORING RAW SWIPE VECTORS FOR CONTINUOUS AUTHENTICATION OF SMARTPHONE USERS

Abstract

Touch based Continuous Authentication (TCA) is a security method that is useful for perpetually validating a user’s identity. In the setting of touchscreen-based smartphones, one type of TCA uses the swipe characteristics of the user to unobtrusively extract hidden patterns to ascertain his or her identity. However, most of the swipe-based TCA methods require this raw touch input data to be pre-processed and scaled to a different form to be of any use for accurate prediction. Further, most of these methods require a user to input multiple swipes before arriving at the authentication decision. This work explores the applicability of methods on the raw swipe data and also attempts to attest users with minimum number of swiping inputs to arrive at the authentication decision at the earliest to minimize damages in case of any unauthorized access. Decent authentication performance is achieved with un-processed and minimal swipe inputs from the user.