Keystroke Authentication Based on Statistical Approach and Changing Input Device for Network Application

摘要

This paper presents techniques to verify the identity of computer users using the keystroke dynamics of computer users login string as characteristic patterns using pattern recognition techniques. This work is a continuation of previous works where other pattern recognition techniques were used. In this work we used hybrid distance measurement. We compute this distance for key hold times as first feature and inter key times as second feature Then we used combined distances for the verification process. We applied several decision methods for final decision boundary and compare results of them. Although it was found that hold times are more effective than inter key times and the best identification performance was achieved by using both time measurements. An verification accuracy of 97% and 99% was achieved for two type of error rates. Another problem has considered in this work is notice to changing input device that is faced in network applications. This point hasn't considered in other work in this area. We present a new method that sense changing input device and then use of this primary result and effect it in next phase of verification process. With this method this biometric verification will be more reliable for general application on wide area network.