Generative Adversarial Attacks on Fingerprint Recognition Systems
Hee won Kwon, Jea-Won Nam, Joongheon Kim, Youn Kyu Lee · 2021
As a recent distribution of mobile and fintech systems, user authentication via biometric recognition has been widely used. Specifically, deep learning-based fingerprint recognition methods became popular due to their high accuracy in fingerprint liveness detection. However, existing deep learning-based approaches contain inherent vulnerabilities such as adversarial attack. In this paper, we propose a new method for generative adversarial attack on fingerprint recognition systems. Our method generates fingerprint images which effectively break existing deep learning-based fingerprint recognition systems while maintaining real-looking appearance of target fingerprints.