PEKS-Based Privacy Preserving Multi-Modal Biometric Identification

Biometric identification is a key component in contemporary authentication systems, traditional architectures however reveal the sensitive templates and are attacked by linkage, reconstruction, inference etc. The existing privacypreserving techniques, such as homomorphic encryption and cancellable biometrics demonstrate security but at the cost of real-time applicability due to their high computational cost, no scalability, and low accuracy. Towards this, in this analysis, a privacy-preserving biometric identity management framework based on Public-Key Encryption with Keyword Search (PEKS) and encrypted fusion of fingerprint, iris and face characteristics are proposed. This system applies optimized pre-processing, encrypted similarity computation, and an efficient encrypted classifier to realize secure 1: N matching without revealing raw biometric information at any stage. We developed a real-time secure access framework for evaluating the encrypted acquisition, matching latency and environmental robustness with fingerprint, iris and face inputs connected to a local PEKS matching engine. The accuracy and precision when performing experiments on CASIA-Iris, FVC2004 and LFW are 98.9%, 98.4% as well as real-time processing of both indoor or outdoor with illumination variance, noisy environment and user head movement. A multi-user load analysis showed that the authentication throughput was stable even when 120 users sent queries simultaneously over encrypted channels, and Differential Privacy further enhanced template protection. The suggested system achieves a good compromise between privacy, accuracy and real time response and is suitable for deploying in access control/border management/smart infrastructure/large scale digital identity systems.