Genetic Algorithm Attack on Minutiae-Based Fingerprint Authentication and Protected Template Fingerprint Systems

Andras Rozsa, Albert E. Glock, Terrance E. Boult; Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR) Workshops, 2015, pp. 100-108

This paper describes a new generic attack against minutiae-based fingerprint authentication systems. The goal of the attack is to construct a fingerprint minutiae template that matches a fixed but unknown reference template. The effectiveness of our attacking system is experimentally demonstrated against multiple fingerprint authentication systems. The paper discusses this attack on two leading privacy-enhanced template schemes and shows it can easily recover high matching score templates. A more general and novel aspect of our work is showing that despite high scores of the attack, the resulting templates do not match the original fingerprint and therefore the underlying data is still privacy protected. We conjecture that the ambiguity caused by collisions from projections/hashing during the privacy-enhanced template production provides for a multitude of minima, which trap attacks in a high-score but non-authentic region.