Secure Biometric Authentication Using Homomorphic Encryption and Fuzzy–Based Similarity Matching

Abstract

This paper presents a secure biometric authentication method that integrates homomorphic encryption, fuzzy-based techniques, and classical cryptographic algorithms to ensure data confidentiality and high recognition accuracy. The proposed approach addresses the limitations of traditional systems, which require access to plaintext biometric templates and are vulnerable to cyber threats. A hybrid model based on the Paillier cryptosystem enables computations directly in the encrypted domain, while fuzzy-based similarity handling accounts for biometric variability. A multi-layered architecture implementing a Defense-in-Depth strategy is proposed, supporting secure feature extraction, encrypted storage, and homomorphic comparison of biometric vectors. Experimental evaluation on the FVC2004 dataset demonstrates high performance, achieving an AUC of 0.995 with reduced FAR, FRR, and EER. The results confirm that the proposed method provides an effective balance between accuracy, security, and computational efficiency.