Fingerprint Recognition Scheme Based on Deep Learning and Homomorphic Encryption

In fingerprint identification, fingerprint features, as a kind of biological feature, are unique, universal, and irrevocable. If it is maliciously attacked, leaked, and tamperedwith, the security of users’ personal fingerprint privacy information will be faced with huge challenges. This paper proposes a fingerprint feature recognition privacy security scheme based on deep learning convolutional neural network and homomorphic encryption algorithm to solve this problem. In this scheme, we add fingerprint classification algorithm to CNN(Convolutional Neural Network) for efficient fingerprint classification and feature extraction. In addition, the BFV homomorphic encryption algorithm is used to encrypt fingerprint feature data and perform matching operations in the ciphertext domain, and fingerprint feature ciphertext database is built to achieve quickly search and matching of feature ciphertext. Finally, we adopt the national secret SM4 and the national secret SM9 algorithms to improve the security of fingerprint signature ciphertext transmission and the ability to resist malicious attacks. The experimental results show that the scheme considers the accuracy of fingerprint identification and overall efficiency and improve the security of fingerprint feature data transmission, storage, and comparison.