Multi-Modal Biometric Authentication System Using Score Fusion Techniques

Abstract

Biometric as an advanced access control method, however, the security can be enhanced through combination of more than one biometric element into one system. This study investigates the enhancement of security in access control systems by implementing a multi-modal biometric authentication system. It explores three biometric combinations: face and fingerprint, face and iris, and fingerprint and iris by using datasets from the CASIA database. The methodology includes biometric image preprocessing, feature extraction using DeepFace (for face), minutiae points (for fingerprints), and Gabor filters (for iris), followed by score-level fusion using weighted average techniques. Experimental analysis reveals that the face-fingerprint combination achieves the highest accuracy of 90.8%, followed by face-iris at 88.8%, outperforming unimodal systems. These results demonstrate the advantage of combining biometric traits for a more reliable and secure authentication system, contributing to the advancement of biometric security technologies.