Ciphertext face recognition system based on secure inner product protocol

Abstract

User privacy data leakage is a major weakness of current plaintext face recognition systems. This article combines Pallier homomorphic encryption algorithm with inner product protocol and face recognition to construct a ciphertext face recognition system based on inner product protocol. Firstly, we integrate the Pallier homomorphic encryption algorithm into the inner product protocol process and design a secure inner product protocol. Through simulation experiments, it can be seen that the time and communication costs of the secure inner product can be reduced to and respectively at the expense of a portion of the applicable scope, where is the vector length. Secondly, we provide specific steps for combining FaceNet face recognition algorithm with secure inner product protocol, then design a ciphertext face recognition system that does not require a trusted third party and analyze its correctness, security and time cost. Through experiments, it can be concluded that the accuracy of plaintext face recognition and ciphertext face recognition on the test set is 98.78% and 98.78%, respectively. The application of Pallier homomorphic encryption algorithm has not affected its recognition performance, which means that the system can provide high accuracy face recognition services while protecting user privacy.