Mobile Gait Match-on-Card Authentication from Acceleration Data with Offline-Simplified Models

Biometrics have become important for authentication on mobile devices, e.g. to unlock devices before using them. One way to protect biometric information stored on mobile devices from disclosure is using embedded smart cards (SCs) with biometric match-on-card (MOC) approaches. Computational restrictions of SCs thereby also limit biometric matching procedures. We present a mobile MOC approach that uses offline training to obtain authentication models with a simplistic internal representation in the final trained state, whereat we adapt features and model representation to enable their usage on SCs. The obtained model is used within SCs on mobile devices without requiring retraining when enrolling individual users. We apply our approach to acceleration based mobile gait authentication, using a 16 bit integer range Java Card, and evaluate authentication performance and computation time on the SC using a publicly available dataset. Results indicate that our approach is feasible with an equal error rate of ∼12% and a computation time below 2s on the SC, including data transmissions and computations. To the best of our knowledge, this thereby represents the first practically feasible approach towards acceleration based gait match-on-card authentication.