A Privacy-Preserving Face Recognition Scheme Combining Homomorphic Encryption and Parallel Computing

Face recognition technology is widely used in various fields, such as law enforcement, payment systems, transportation, and access control. Traditional face authentication systems typically establish a facial feature template database for identity verification. However, this approach poses various security risks, such as the risk of plaintext feature data stored in cloud databases being leaked or stolen. To address these issues, in recent years, a face recognition technology based on homomorphic encryption has gained attention. Based on homomorphic encryption, face recognition can encrypt facial feature values and achieve feature matching without exposing the feature information. However, due to the encryption, face recognition in the ciphertext domain often requires considerable time. In this paper, we introduce the big data stream processing engine Flink to achieve parallel computation of face recognition in the ciphertext domain based on homomorphic encryption. We analyze the security, accuracy, and acceleration of this approach. Ultimately, we verify that this approach achieves recognition accuracy close to plaintext and significant efficiency improvement.