AI-BASED FEATURETRANSFORMATION TO CANCELLABLE FINGERPRINT AUTHENTICATION WITH BLOCKCHAIN

Abstract

This paper provides a secure and privacy-sensitive biometric authentication system that combines deep learning, cancellable biometrics, and blockchain technology. The discriminative fingerprint features are extracted by a convolutional neural network and undergo deterministic strong cancellable transformation as well as irreversibly hashed with SHA-256 to avoid reconstruction. The hashed templates are made and checked with an Ethereum smart contract and hence data immutability and decentralized trust. The system also removes the threat of biometric leakage and replay attack as the unique and revocable templates are regenerated and original images are not stored. Experimental evidence proves the solidity of the framework in the non-invertibility of features and accuracy of authentication with the efficiency to process the features in real time. This paper points to a scalable, transparent, and tamper-resistant model of biometric identity management that can be used in next-generation digital security systems.