Multimodal smartphone user authentication using touchstroke, phone-movement and face patterns

Recently, researches have shown to employ implicit behavioral biometrics via built-in sensors (e.g., gyroscope) for user identification on smartphones. The majority of prior studies are based on unimodal systems, which suffer from low accuracy, spoofing and lower usability. In this paper, we present an unconstrained and implicit multimodal biometric system for smartphones using touchstroke, phone-movement and face patterns. The proposed framework authenticates the user by taking silently into account micro-movements of the phone1, movements of the user's finger during typing on the touchscreen, and user's face features. We also collected a mobile multimodal dataset of touchstroke and phone-movement patterns in the wild from 95 subjects. Preliminary experimental analysis on accuracy and usability show promising results.