Accelerating Authentication Using Locality-Sensitive Hashing in a Homomorphic Encryption-Based Face Recognition System

Homomorphic encryption has garnered significant attention in addressing the critical need for privacy and security in facial recognition systems. Although this technique enhances data confidentiality, its practical application remains challenging owing to its substantial computational overhead and significant processing delays. This paper proposes an approach integrating locality-sensitive hashing (LSH) to accelerate the authentication process. LSH is a computational technique designed to identify similar data points in high-dimensional spaces. We utilize LSH to pre-classify similar facial features, effectively narrowing the search space and reducing the computational cost during authentication. This approach delivers significantly improved processing speeds and enhanced scalability to accommodate larger databases. In an experiment with 250 registered users, the conventional system had an average response time of 4.957±0.256 s, whereas the LSH-enhanced system reduced it to 0.307±0.135 s. This research contributes to improving the practicality of homomorphic encryption-based facial recognition systems.