Towards quantifying the entropy of fingervein patterns across different feature extractors

This paper makes a first attempt at quantifying the entropy of fingervein patterns that have been extracted using three different state-of-the-art feature extractors, on two publicly-available fingervein databases. We show that the resulting entropy is dependent upon both the feature extractor and database, implying that a universal estimate of fingervein entropy would be misleading. We also discuss how our entropy results can be applied towards more meaningful evaluations of the security and privacy of fingervein template protection schemes. Our open-source implementation ofthe entropy estimation on a publicly-available fingervein recognition system will help the research community to both validate our findings and build upon our work.