Improvement of keystroke data quality through artificial rhythms and cues

Abstract

Keystroke dynamics based user authentication (KDA) can achieve a relatively high performance if a fairly large number of typing patterns are available. It is almost always the case that KDA is combined with password based authentication. Users are often required to change their passwords. When a user changes one's password, however, only a handful of new patterns become available. In a mobile situation, moreover, very short passwords are used. Under such a circumstance, the quality of data becomes important. Recently, artificial rhythms and cues were proposed to improve the quality of data. In this paper, we verify the effectiveness of artificial rhythms and cues through hypotheses tests using the data from 25 users under various situations. The experimental results show that artificial rhythms increase the uniqueness while cues increase the consistency.