Dynamic authentication on mobile devices: evaluating continuous identity verification through swiping gestures

Biometrics, the science of identifying individuals based on unique physiological and behavioral traits, has witnessed widespread adoption in recent years due to its applications in security, access control, and authentication. Behavioral biometrics, which leverage unique behavioral patterns, offer a non-intrusive and user-friendly approach to identity verification. Swiping gestures, a fundamental interaction mechanism on mobile devices, hold significant promise for continuous verification. This paper delves into the domain of behavioral biometrics, specifically focusing on the utilization of swiping gestures for continuous identity verification on mobile devices. Unlike discrete verification methods, continuous verification offers an ongoing assessment of an individual's authenticity, aligning with the pace of modern interactions while enhancing security. We evaluate various Machine Learning one-class classifiers, including a deep learning model, to evaluate continuous verification systems while using a huge real-world and publicly available dataset. Our results shows that the used deep learning model performs well in all scenarios of test compared to traditional classifiers. A good value of the Equal Error Rate equal to 0.20% is achieved using the deep learning model. The used models in this paper can be downloaded from this link: https://github.com/AnassSej/Dynamic-Authentication-Swipes/.