Securing Biometric Data: Fully Homomorphic Encryption in Multimodal Iris and Face Recognition

Abstract: Multimodal biometric systems have gained popularity for their enhanced recognition accuracy and resistance to attacks like spoofing. This research explores methods for fusing iris and face feature vectors and implements robust security measures to protect fused databases and conduct matching operations on encrypted templates using fully homomorphic encryption (FHE). Evaluations on the QFIRE-I database demonstrate that our method effectively balances user privacy and accuracy while maintaining a high level of precision. Through experimentation, we demonstrate the effectiveness of employing FHE for template protection and matching within the encrypted domain, achieving notable results: a 96.41% True Acceptance Rate (TAR) for iris recognition, 81.19% TAR for face recognition, 98.81% TAR for iris fusion (left and right), and achieving a 100% TAR at 0.1% false acceptance rate (FAR) for face and iris fusion. The application of FHE presents a promising solution for ensuring accurate template matching while safeguarding user privacy and mitigating information leakage.