Hybrid Behavior-based Biometric Authentication Systems (HBAS)

Abstract

An idea of proposing hybrid security based on behavioural biometrics is displayed in this paper, on the two bases of tapping and swiping. Tapping and swiping are the main focus as they are the easiest actions to perform on a mobile device. Actions and mannerisms performed by either attacker and/or the user are explained alongside the results of said actions and mannerisms. An attacker will haphazardly try to brute force the motion while the owner will perform his action consistently. Then, a training set is done where an attacker and an owner’s training sets are obtained and set, to determine which classifiers work best in determining the confusion matrix from a range of training set values with Hoeffiding Tree and Naïve Bayes Multinomial classifiers performing the best over the three sets of training values provided.