Effectiveness of authenticating users with randomly constructed fingerprint templates

In recent years, using fingerprints to authenticate users in physical and logical access control has been gaining popularity. Thus protecting fingerprint data becomes an important issue. One of the ideas proposed in literature to safeguard fingerprint data is to construct secure fingerprint template from fingerprint image(s). One requirement for this proposal is that the template can be utilized to authenticate its owner but an impostor cannot use it to reconstruct the original fingerprints. How to satisfy this requirement is still an ongoing research problem. Partial fingerprints are more readily available than full fingerprints. Application of 3D finger scanner renders fast capturing of multiple partial fingerprints without any distortion. Recognition based on partial fingerprints from different fingers can significantly increase the number of available fingerprint templates. A secret-key based randomized selection of multiple partial fingerprints from multiple fingers can protect the privacy of fingerprint data. In this paper we propose a novel approach to constructing fingerprint templates from multiple partial fingerprint images. The randomized construction process makes reverse engineering more resource consuming. Our testing results show that individual partial fingerprints from different parts of a finger do not match with each other. By enrolling a combined template from multiple partial fingerprints, authentication can be done successfully. The matching score exponentially increase with the number of minutiae in a fingerprint. The randomly constructed template from multiple original templates does not cause an increase in false recognition rate (FRR), compared with the matching results of the original template.