A Novel Fingerprint Biometric Cryptosystem Based on Convolutional Neural Networks

Modern access controls employ biometrics as a means of authentication to great extent. For example, is used an mechanism implemented on commercial devices such smartphones and laptops. This paper presents fingerprint biometric cryptosystem based the fuzzy commitment scheme convolutional neural networks. One its main contributions novel approach automatic discretization texture descriptors, entirely network, designed generate fixed-length templates. By converting templates into binary domain, we developed that can be in key-release systems or template protection matching systems. The problem data variability marginalized by applying secure block-level Bose–Chaudhuri–Hocquenghem error correction codes, resistant statistical-based attacks. evaluation shows significant performance gains when compared other texture-based cryptosystems.