Finger Vein Presentation Attack Detection with Optimized LBP Variants

Abstract

Finger vein-based authentication systems have been proven to be promisingly accurate in identifying a person. However, the system is still highly vulnerable from presentation attack. Presentation attack is one of the most commonly found attacks in typical biometrics systems. A printed finger vein image could be used to bypass the system with ease. Various presentation attack detection methods based on texture and liveness analysis have been presented to encounter such issue. In this paper, our aim is to apply hyper-parameters tuning on Local Binary Pattern to gain the best features set for presentation attack detection in finger vein recognition. Using an automated hyper-parameter tuning approach, we find a set of optimized parameters which are able to extract the best features for presentation attack detection. Experiment results demonstrate that the proposed method is able to yield a significant high accuracy in distinguishing genuine images from fake images.