One-Class Models for Continuous Authentication Based on Keystroke Dynamics

摘要: In this paper we discuss an applied problem of continuous user authentication based on keystroke dynamics. It is important for a model to discover new intruders. That means don’t have the samples such intruders training phase. leads us necessity using one-class models. review some popular feature extraction, preprocessing and classification methods problem. We propose approach reduce dimensionality space two-sample Kolmogorov-Smirnov test investigate how quantile-based discretization technique can improve models’ performance. present two algorithms, which not been used dynamics before: Fuzzy kernel-based classifier Random Forest Regression classifier. conduct experimental evaluation proposed approach.