Adaptive Biometric Systems using Ensembles

With the increased availability of online services, enhanced authentication mechanisms-including biometric systems-are necessary. However, recent studies show that biometric features can change. Consequently, recognition performance can be affected over time. Adaptive biometric systems that can automatically adapt the biometric reference have been proposed to deal with this problem. Frequently, these systems use query samples classified as genuine to adapt the biometric reference. Despite good results, there are concerns regarding their robustness. This article investigates using an ensemble of classifiers to increase these systems robustness. Ensembles can improve the recognition performance of decision models, providing a more stable classification decision. The authors explore questions regarding the application of ensembles to adaptive biometric systems using one-class classification algorithms, and offer a proposal to automatically adapt the meta classifier over time.