Quality-Based Score-level Fusion for Secure and Robust Multimodal Biometrics-based Authentication on Consumer Mobile Devices

Biometric authentication is a promising approach to access control in consumer mobile devices. Most current mobile biometric authentication techniques, however, authenticate people based on a single biometric modality (e.g., iPhone 6 uses only fingerprints), which limits resistance to trait spoofing attacks and ability to accurately identify users under uncontrolled conditions in which mobile devices operate. These challenges can be alleviated by multimodal biometrics or authentication based on multiple modalities. Therefore, we develop a proof-of-concept mobile biometric system which integrates information from face and voice using a novel score-level fusion scheme driven by the quality of the captured biometric samples. We implement our scheme on the Samsung Galaxy S5 smartphone. Preliminary evaluation shows that the approach increases accuracy by 4.14% and 7.86% compared to using face and voice recognition individually, respectively. Keywords–Multimodal biometrics; quality; score-level fusion; mobile