A Voronoi neighbor structure-based fingerprint cryptosystem using minutiae descriptors and error-correcting codes

Minutiae local structure-based fingerprint cryptosystems provide an innovative solution for identity authentication as well as fingerprint protection. However, there are two challenges that need to be addressed. First, minutiae local structure-based fingerprint matching algorithms only use relative local minutiae features (rotation and translation invariant), which are less distinctive compared with both global and local features used in conventional fingerprint matching algorithms. The lack of distinctive features will lead to higher FAR because different fingerprints are very likely to have similar minutiae local structures. Besides, similarity measures for fingerprint templates are not directly applicable to existing bio-cryptosystem constructions. Though there has been much work using quantization to handle nonlinear distortion and transform original templates to bio-cryptosystem-applicable formats, the quantization process itself is by no means a similarity-preserving transformation. In specifics, features that are originally close to/far from each other may be mapped to different/same quantization intervals. In this paper, we propose a Voronoi neighbor structure-based fingerprint cryptosystem using minutiae descriptors and error-correcting codes. The use of minutiae descriptors significantly increases the distinctiveness of minutiae local structures while error-correcting codes effectively rectify nonlinear distortion and preserve fingerprint similarity to a great extent. As a result, the proposed system far outperforms some other minutiae local structure-based fingerprint cryptosystems over the publicly available fingerprint databases.