ITSME: Multi-modal and Unobtrusive Behavioural User Authentication for Smartphones

In this paper, we propose a new multi-modal behavioural biometric that uses features collected while the user slide-unlocks the smartphone to answer a call. In particular, we use the slide swipe, the arm movement in bringing the phone close to the ear and voice recognition to implement our behaviour biometric. We implemented the method on a real phone and we present a controlled user study among 26 participants in multiple scenario’s to evaluate our prototype. We show that for each tested modality the Bayesian network classifier outperforms other classifiers (Random Forest algorithm and Sequential Minimal Optimization). The multimodal system using slide and pickup features improved the unimodal result by a factor two, with a FAR of 11.01% and a FRR of 4.12%. The final HTER was 7.57%.