Independent performance evaluation of pseudonymous identifier fingerprint verification algorithms

Abstract

To develop a biometric product the system evaluator is in many cases the same entity as the system developer. Besides, the test data used for evaluation could be also collected from the same entity. Such databases are in most cases not public ones and thus the test results are hard to be independently verified. This work reports an independent evaluation on fingerprint verification accuracy performance within the context of the EU-FP7 project TURBINE. This paper represents the algorithm developers and evaluators with independent entities. The algorithm developers are assumed not to have the right to accessing the test database consisting of more than 70000 fingerprint samples from 100 subjects captured by 6 fingerprint sensors during months. These controls provide an unbiased and trustworthy performance evaluation. This work evaluates biometric performance on the protected reference level, in addition to the sample- or minutiae- based references. The biometric references evaluated are fingerprint pseudonymous identifiers generated by the template protection algorithms. The performance evaluation results in this work are generated from different template protection algorithms at the pseudonymous identifier levels. © 2013 Springer-Verlag.