Utility of Quality Metrics in Partial Fingerprint Recognition

Abstract

The performance of a fingerprint recognition system depends on the amount of discriminating data available in a fingerprint. The partial fingerprints do not contain enough details for successful matching. Partial fingerprints does not only degrades the matching performance but also introduces a security flaw in an authentication system. The probability of attacking many users with the help of dictionary attack is much higher when the fingerprints are partial. For reliable fingerprint matching and reducing the security flaw, it is necessary to check whether the fingerprint is partial or not. Fingerprint Quality metric plays an important role in assigning quality value to each fingerprint according to the content available in it. Depending on the quality value, the recognition system needs to take the decision on whether to allow, enhance, or re-acquire the fingerprint image. The paper assesses the ability of a fingerprint image quality metric to detect partial fingerprints as low-quality fingerprints. Extensive evaluation of the ten fingerprint image quality methods is performed to check their performance on partial fingerprints in terms of Utility. A new partial fingerprint dataset is prepared by cropping the fingerprints in FVC 2004 DB1A dataset to check the ability of quality metrics in handling the partial fingerprints. To calculate the match scores, two minutia-based matchers, NIST-Bozorth3 and K-Plet are used. The experimental results in this research shows that NFIQ 2.0 and Gabor based method are good at detecting partial fingerprints by assigning them low quality values.