Parallel Prints: Generating Realistic Cancelable Fingerprint Templates

Rishabh shukla, Harkeerat Kaur, Isao Echizen; Proceedings of the Winter Conference on Applications of Computer Vision (WACV) Workshops, 2025, pp. 1449-1458

Fingerprints are widely recognized as a crucial method for uniquely identifying individuals. Nevertheless the utilization of fingerprints in an online platform presents a noteworthy concern for privacy as the original fingerprints are constantly vulnerable to various kinds of attacks thefts forgery unintended storage and processing by third parties. The use of original fingerprints for authentication is riddled with concerns surrounding privacy and security. To address this we introduce artificial intelligence techniques to transform the original unique fingerprint into a reliable natural looking template that can be stored and used for authentication so that the need for decoding at the end of a third party is eliminated. The study proposes a novel approach to creating extremely realistic fingerprint templates while also providing the ability to revoke and cancel stolen fingerprints in the event of an attack. We used the various publicly available datasets during the training and testing phases. We tested the quality of generated data throughout the testing phase where the system performance is reported at EER of 0.06% and AUC of 0.99.