A Novel Key Generation Scheme for Biometric Cryptosystems Using Fingerprint Minutiae

In this paper, a novel method of generating cancelable cryptographic key using local features from fingerprint minutiae points has been proposed. Local features from input fingerprint image are determined and minutiae points are extracted using Gabor filter. Then a template matrix is generated from extracted minutiae points. Two directed acyclic graphs (DAG) are generated from the template matrix taking minutiae coordinates as edges. Binary values are assigned to each edge based on local features of the minutiae points-orientation, frequency and phase. Then for each DAG, the proposed algorithm selects each node as source and determines the longest reachable path, traverses the path and generates a binary string. Thus, two binary strings are generated from the two DAGs. Finally the proposed algorithm performs a simple rotation and combination operation on the two strings and generates four cryptographic keys. We have performed experiments on existing benchmark fingerprint databases and based on security analysis of our method we conclude that our proposed method is a simple and efficient method that generates four cancelable cryptographic keys and the proposed process has very strong irreversible property.