Reconstruction of Fingerprints from Minutiae Using Conditional Adversarial Networks

Abstract

Fingerprint recognition systems have been known to be exposed to several security threats. Those are fake fingerprints, attacking at communication channels and software modules, and stealing fingerprint templates from database storages. For a long time, stolen templates are not seriously investigated because it was believed that fingerprint templates did not reveal the original fingerprints used to extract the templates. However, recent studies have proved that a fingerprint can be reconstructed from its minutiae, although the reconstructed fingerprints may have many spurious minutiae and unnatural patterns. This paper proposes an algorithm based on conditional generative adversarial networks (conditional GANs) to reconstruct fingerprints from sets of minutiae. The fingerprints generated by the proposed networks are very similar to the real fingerprints and can be used to fool fingerprint recognition systems. The acceptance rates of the generated fingerprints range from 42% to 98%, depending on the features and security levels used in the matching algorithms.