POP-FA: Paillier and Order-Preserving encryption-based Facial Authentication
Yatish Dubasi, William Andrew Burroughs, Qinghua Li · 2024
Facial recognition access control systems are becoming increasingly prevalent, offering a convenient and secure way to control access to physical and digital spaces. However, these systems also raise significant privacy concerns. When users enroll in server-based face recognition access control (SFRAC) systems, their facial data is captured, processed, and stored on a server. The data at the server can be vulnerable to unauthorized access or breaches out of users' control, potentially leading to misuse or identity theft. This paper proposes Paillier and Order-Preserving encryption-based Facial Authentication (POP-FA), a novel framework that addresses these privacy concerns in SFRAC systems. POP-FA utilizes a combination of Paillier encryption and Order-Preserving Encryption (OPE) to protect facial privacy while still allowing the computation of similarity between encrypted facial embeddings to authenticate users. This ensures that users’ facial data remains confidential from the server throughout the enrollment and authentication processes. Our comprehensive analysis and experiments demonstrate that POP-FA achieves high accuracy in facial recognition and meanwhile maintains strong resistance to impersonation attacks. Additionally, POP-FA exhibits high processing efficiency, making it a viable solution for real-world SFRAC deployments.