Novelty detection approach for keystroke dynamics identity verification

Abstract

Password is the most widely used identity verification method in computer security domain. However, due to its simplicity, it is vulnerable to imposter attacks. Keystroke dynamics adds a shield to password. Discriminating imposters from owners is a novelty detection problem. Recent research reported good performance of Auto-Associative Multilayer Perceptron(AaMLP). However, the 2-layer AaMLP cannot identify nonlinear boundaries, which can result in serious problems in computer security. In this paper, we applied 4-layer AaMLP as well as SVM as novelty detector to keystroke dynamics identity verification, and found that they can significantly improve the performance. © Springer-Verlag 2003.