Multi-Layered Biometric Obfuscation (MLBO): Enhancing Security in Facial Recognition Systems

Abstract

As facial recognition technologies become ubiquitous, the need for robust biometric data protection is more critical than ever. This paper introduces Multi-Layered Biometric Obfuscation (MLBO), a novel framework designed to enhance the security and privacy of facial recognition systems. By combining neural key generation, frequency-domain perturbation, and secret projection, MLBO achieves the three core properties of cancelable biometrics: revocability, diversity, and non-invertibility.Experimental evaluation on the ORL and Yale face databases demonstrates that MLBO provides strong biometric template protection while maintaining a competitive Equal Error Rate (EER) of approximately 6%. These results confirm that MLBO offers a practical and well-balanced solution for deploying secure, privacy-preserving facial recognition systems.