Analysing a Vein Liveness Detection Scheme

Abstract

We examine a previously published liveness detection method for guarding against presentation attacks on vein recognition systems which employs motion magnification on video frames, and develop three new attacks that circumvent the proposed protective scheme. The first pair of attacks are direct attacks or presentation attacks, and involve presenting a fake sample with rhythmic motions to the biometric system. The third attack is an indirect attack that feeds the biometric system a synthetic video signal designed to circumvent the liveness detection scheme. Results show that the analysed liveness detection system must not be used as a standalone technique. We conclude by recommending improvements to the analysed scheme to harden against attacks of the kind we presented and to avoid having to combine it with other presentation attack detection techniques.