Continuous Authentication Using One-class Classifiers and their Fusion

摘要

While developing continuous authentication systems (CAS), we generally assumethat samples from both genuine and impostor classes are readily available.However, the assumption may not be true in certain circumstances. Therefore, weexplore the possibility of implementing CAS using only genuine samples.Specifically, we investigate the usefulness of four one-class classifiers OCC(elliptic envelope, isolation forest, local outliers factor, and one-classsupport vector machines) and their fusion. The performance of these classifierswas evaluated on four distinct behavioral biometric datasets, and compared witheight multi-class classifiers (MCC). The results demonstrate that if we havesufficient training data from the genuine user the OCC, and their fusion canclosely match the performance of the majority of MCC. Our findings encouragethe research community to use OCC in order to build CAS as they do not requireknowledge of impostor class during the enrollment process.