Unlocking a phone verifies its user at a single moment. This paper explores how authentication can continue during use by combining touch gestures and accelerometer data. We implemented an Android system that records these signals and uses a Hidden Markov Model to distinguish the authenticated user from others. The research connects mobile application development, behavioural biometrics, and machine learning.
From research to software
The supporting repository includes an Android data recorder, Python model-training notebooks, and an Android verification application. It allows the implementation to be explored alongside the publication.
Citation
Reichinger, D.; Sonnleitner, E.; Kurz, M. Continuous Mobile User Authentication Using Combined Biometric Traits. Applied Sciences 2021, 11(24), 11756. https://doi.org/10.3390/app112411756.