SPIN: Leveraging Fuzzy Commitment and LDPC Sum-Product Decoder for Key Generation from Face

Yafei Liu, Ying Zhou, Weiyu Zhou, Hui Zhang, Yufang Dong, Xingbo Dong, Zhe Jin

Abstract: Biometric authentication, with its unique and convenient features, is gaining popularity as a secure access control method for diverse systems and services. However, the inherent ambiguity of biometric data poses challenges when integrating it with cryptographic systems that require 100% accuracy, such as personal information number generation. To address this, we propose SP\(^2\)IN, a face bio-cryptosystem that utilizes a Low-Density Parity-Check Sum-Product decoder and fuzzy commitment. In our innovative face biometric cryptosystem, we employ fuzzy commitment for secure key extraction from biometric input, ensuring the protection of sensitive facial information without compromising the privacy of the raw data. To tackle facial biometric noise, we use a Low-Density Parity-Check Sum-Product decoder for error correction against variations. Our system was rigorously tested on public face datasets (LFW, AgeDB, CALFW), showcasing outstanding recognition rates: 99.43% (LFW), 90.43% (CALFW), 92.63% (AgeDB30).