Efficient and Practical Privacy-Preserving Face Recognition Protocol

With the rapid development and wide application of artificial intelligence technology, face recognition has been an important research topic in the field of artificial intelligence. The widespread use of biometrics in daily life has raised public concerns about the misuse and privacy of facial images from users. Therefore, there is an urgent need to address the preservation of privacy of facial data. Under this premise, many face recognition privacy preserving protocols have emerged, which are helpful to ensure that sensitive user face data are not abused or leaked. However, most existing privacy-preserving facial recognition protocols require extensive computation of ciphertexts during execution to ensure the security of the corresponding data. However, this reduces the efficiency of the protocol. In addition, phases such as adding users and deleting users are overlooked, which makes the system not practical enough. To address the problem of protecting the privacy of face data while ensuring its efficiency and practicality in real-world scenarios, we propose an privacy-preserving protocol based on homomorphic encryption and a dynamic sorting algorithm trained using the AlphaDev model, which improves the computational efficiency of the protocol under the premise of guaranteeing the security, and adds a phase of adding or subtracting users to ensure its practicality. Finally, simulation experiments demonstrate that our scheme can improve the efficiency and practicality of face recognition while preserving privacy.