A fingerprint cryptosystem based on minutiae phase spectrum

Public acceptance of biometric technology is hindered by security and privacy concerns regarding leakage of biometric templates. Biometric cryptosystems alleviate this problem by storing a secure sketch that is typically derived by binding the template with a cryptographic key. However, designing fingerprint cryptosystems is difficult because fingerprint matching is usually based on unordered sets of minutia features having large intra-user variations. To overcome this challenge, we propose a novel minutiae representation known as the Binarized Phase Spectrum (BiPS), which is a fixed-length binary string obtained by quantizing the Fourier phase spectrum of a minutia set. We secure the BiPS representation using a fuzzy commitment scheme employing turbo codes. We also propose a technique for aligning fingerprints based on the focal point of high curvature regions. The proposed system achieves a FNMR of 16.2% and 12.6% on FVC2002-DB1 and DB2 databases, respectively, at zero FMR.