Keystroke and swipe biometrics fusion to enhance smartphones authentication

Abstract

Several authentication techniques are required to preserve smartphone users' privacy. Part of these authentication mechanisms are based on Keystroke Dynamics (KD) or Swipe Dynamics (SD); however, these mechanisms have performance challenges due to the following; limited capability in handling user behavioral variations, inefficient feature extraction, and alternate usages that are not restricted to a specific method typing or swiping. This work presents an improved smartphone continuous authentication model by integrating free text-based KD and SD. The proposed model adopts feature-level fusion by concatenating free-text KD and SD features. This Feature level fusion was evaluated based on a comprehensive and benchmark dataset and Random Forest (RF) classifier. Results have confirmed the proposed model's performance, in which accuracy was 99.98%, with the lowest Equal Error Rate (EER) rate of 0.02% based on multi-class classification.