Graph based secure cancelable palm vein biometrics

Palm vein is a promising vascular biometric modality that offers contactless and accurate authentication. The complex nature of the vein pattern provides uniqueness to each individual which helps in efficient individual authentication. However, most of the existing palm vein pattern-based works extract point-specific features like bifurcation and endpoints. These methods overlook the unique connections between the vein points. Graphs can help in representing the entire vein structure. However, a direct spatial graph representation of the vein pattern poses template security vulnerabilities. Therefore, a graph representation that uses the structure information, while providing template security is required. Towards this, in this paper, we propose a graph representation for palm veins with vein pattern blocks as nodes and the vein connection between neighbouring blocks as edges. The vein pattern is partitioned into blocks based on random numbers generated using a user-specific key which guarantees template security and cancelability. The method of block generation and edge connection gives uniqueness to this graph. This approach requires no additional processing for providing template security as the feature representation automatically ensures it. Here, even the compromise of key does not affect template security. Extensive experiments with three publicly available datasets prove the efficacy of the method.