Deep learning and encryption algorithms based model for enhancing biometric security for artificial intelligence era

Abstract

The significance of facial recognition in the era of artificial intelligence lies in its utilization of facial features as a type of biometric characteristic possessing uniqueness and irreversibility. However, exposing these features to attacks, tampering, or unauthorized disclosure poses considerable threats to user privacy and security. A privacy and security solution based on deep learning and encryption algorithms is proposed to tackle this issue. This solution employs the FaceNet deep learning algorithm to extract facial features efficiently. The combination of biometric feature blurriness and cryptographic system precision is achieved, utilizing the CKKS fully homomorphic encryption algorithm for operations in the ciphertext domain of facial recognition. The SM4 algorithm is used to enhance the resilience of facial feature ciphertext against malicious attacks. By leveraging the properties of symmetric ciphers, a balance is achieved between security and computational efficiency. The management of the symmetric key used in the SM4 algorithm is conducted through the employment of the SM9 asymmetric encryption algorithm. Experimental results and analysis demonstrate that the proposed solution enhances the security of data transmission, storage, and comparison without compromising the accuracy and efficiency of facial recognition.