Privacy preserving unique robust and revocable passcode generation from fingerprint data

Abstract

This research explores generating passcodes from fingerprint images. The investigation unfolds through a three-step process: (a) fixed length feature vector generation from minutia features, (b) stable binary feature vector generation from the fixed length feature vector, and (c) passcode generation from the stable binary feature vector. The main research objectives of this work are: (1) how a unique and robust binary pattern can be generated from a fingerprint image, (2) from this binary bit pattern, how a passcode can be generated satisfying non-linkable and revocable properties, and (3) how an attacker cannot guess the source biometric given a passcode thus preserving the privacy of the fingerprint data. The generated passcode can be applied in many applications, such as unique identity generation for authentication without enrollment, encryption key generation for network security, remote authentication protocol or distributed systems, data storage security, digital wallet, etc. The proposed approach has been validated with FVC2002 and FVC2004, and results show impressive genuine acceptance rates of 99.31% and 99.25%, with 0% false acceptance rates. Further, the generated passcodes pass NIST and Diehard randomness tests, substantiating the potential key generation technique with high intra-similarity and low inter-similarity.