Fusion of face and periocular information for improved authentication on smartphones

Authentication on smartphones for high valued transactions in e-banking applications require authentication methods that are both more robust and more convenient than simple PIN based authentication. Although a two factor authentication approach can increase the system security, the second factor has to be robust enough not to be compromised by simple attacks. Face and periocular based biometrics have been explored in many works for authentication on smartphone. Multi-modal biometrics always have an edge over the unimodal biometrics. In this work, we explore various multi-modal biometrics employing feature level and score level fusion to improve the biometric performance of the system. An extensive number of experiments are conducted using two different devices - Samsung Galaxy S5 smartphone and Samsung Galaxy Note 10.1 tablet. The database consisting of 46 subjects has provided a performance with best Genuine Match Rate (GMR) of 95.52% for smartphone and 96.65% GMR for tablet. The obtained scores for feature level and score level fusion advocates the use of a fused approaches over a unimodal biometric authentication on the smartphones.