Design of Alignment-Free Cancelable Fingerprint Templates with Zoned Minutia Pairs

Cancelable fingerprint templates effectively protect original fingerprint data by revoking a compromised template and reissuing a new one. Alignment-free cancelable templates require no image pre-alignment and therefore do not suffer from inaccurate singular point detection. In this paper we propose to construct local minutia structures through zoned minutia pairs in the design of alignment-free cancelable fingerprint templates. Because the structures formed by zoned minutia pairs are truly local, they are more discriminating, which benefits fingerprint matching, leading to good recognition performance. We also apply a simple yet effective measure to mitigating the risk of the attack via record multiplicity (ARM). The proposed method features a partial discrete Fourier transform-based non-invertible transformation, which is compact and computationally efficient. The new cancelable templates meet the requirements of diversity, revocability, non-invertibility and performance. Evaluated over four public databases, the proposed alignment-free cancelable fingerprint templates exhibit superior performance with the Equal Error Rate in the lost-token scenario being 0.19% for FVC2002 DB1, 1% for FVC2002 DB2, 4.29% for FVC2002 DB3 and 9.01% for FVC2004 DB2.