Abstract:Face verification is a well-known image analysis application and is widely used to recognize individuals in contemporary society. However, most real-world recognition systems ignore the importance of protecting the identity-sensitive facial images that are used for verification. To address this problem, we investigate how to implement a secure face verification system that protects the facial images from being imitated. In our work, we use the DeepID2 convolutional neural network to extract the features of a facial image and an EM algorithm to solve the facial verification problem. To maintain the privacy of facial images, we apply homomorphic encryption schemes to encrypt the facial data and compute the EM algorithm in the ciphertext domain. We develop three face verification systems for surveillance (or entrance) control of a local community based on three levels of privacy concerns. The associated timing performances are presented to demonstrate their feasibility for practical implementation.

What problem does this paper attempt to address?

The core problem that this paper attempts to solve is: to realize an efficient and accurate face verification system on the premise of ensuring personal privacy and data security. Specifically, the paper focuses on how to perform facial feature extraction and verification in the encrypted domain to prevent sensitive facial images from being imitated or leaked. ### Problem Definition With the development of modern society, identity verification has become increasingly important, especially when accessing certain services or obtaining specific rights (for example, when users log in to websites or use fingerprint unlocking on mobile phones). As a common method of identity verification, face recognition has been widely applied. However, most of the existing face recognition systems ignore the importance of protecting the facial images used for verification. These images contain personal sensitive information and are easily misused or leaked. ### Main Contributions of the Paper 1. **Convolutional Neural Network (CNN) Model Based on DeepID2**: - Use the DeepID2 CNN model to extract features from facial images. - Apply the EM algorithm to verify whether two given images are from the same identity. 2. **Application of Homomorphic Encryption**: - In order to protect personal privacy, apply a homomorphic encryption scheme to encrypt facial image features and execute the EM algorithm in the ciphertext domain. - In this way, face verification can be carried out without decryption, thus protecting the security of facial images. 3. **Development of Practical Application Systems**: - Develop three face verification systems for community surveillance or entrance control, corresponding to different privacy protection levels respectively. - Demonstrate the feasibility and time performance of these systems through experiments, proving their potential in practical applications. ### Main Challenges - Analyzing data in the encrypted domain and maintaining high verification accuracy is one of the main challenges in this paper. - It is necessary to ensure the efficiency and accuracy of the system while guaranteeing data privacy. ### Conclusion By combining deep - learning and homomorphic encryption technologies, this paper proposes a new method that can not only protect personal privacy but also efficiently perform face verification. This method not only improves the security of face verification but also provides new ideas for future privacy - protection research.

A Machine Learning-Based Secure Face Verification Scheme and Its Applications to Digital Surveillance

DepthFake: Spoofing 3D Face Authentication with a 2D Photo

An Intelligent Surveillance System Base on Accelerated Human and Face Detection

Design and Implementation of Identity Verification Software Based on Deep Learning

Realization of a Hybrid Locally Connected Extreme Learning Machine With DeepID for Face Verification

Multi-modal Spoofing Attacks on 3D Face Liveness Detection Via a Single 2D Photo

Face Detection and Recognition from Distance Based on Deep Learning

Convolutional-based variational autoencoders for face privacy protection in video surveillance

Verifiable Facial De-Identification in Video Surveillance

Deep Secure Encoding: An Application to Face Recognition

UniID: Spoofing Face Authentication System by Universal Identity

Time-Aware Face Anti-Spoofing with Rotation Invariant Local Binary Patterns and Deep Learning

Privacy-preserving Face Detection Based on Linear and Nonlinear Kernels.

Enhancing societal security: a multimodal deep learning approach for a public person identification and tracking system

A Face-Unlock Screen Saver by Using Face Verification Based on Identity-Specific Subspaces

Face Detection for Privacy Protected Images

Application of Face Recognition Method Under Deep Learning Algorithm in Embedded Systems

People Taking Photos That Faces Never Share: Privacy Protection and Fairness Enhancement from Camera to User

Secure Dual Network for Reversible Facial Image Anonymization Through the Latent Space Manipulation

Design and implementation of a surveillance camera system with face recognition functionality

Identifiable Face Privacy Protection Via Virtual Identity Transformation