Novel impostors detection in keystroke dynamics by support vector machine

To detect the novel impostors whose data patterns have never been learned previously in keystroke dynamics, two solutions are proposed in this paper. Unlike most other research in keystroke dynamics, this paper surveys the performance tradeoff and time consumption, which are valuable for practical implementation, of the solutions. Besides, it is our intention to attempt verifying computer users' identities based on pure numeric password which is more difficult than verification of any other kinds of passwords.