Skip to Main Content
Keystroke dynamics exhibit people's behavioral features which are similar to hand signatures. A major problem hindering the large scale deployment of this technology is its high FAR (false acceptance rate) and FRR (false rejection rate). A significant progress, in terms of improving the FAR and FRR performance, has been made by the work of Gunetti and Picardi (2005). However, their identification based authentication suffers a severe scalability issue as it needs to verify the input with every training sample of every user within the whole database. In this paper, a k-nearest neighbor approach has been proposed to classify users' keystroke dynamics profiles. For authentication, an input will be checked against the profiles within the cluster which has greatly reduced the verification load. Experiment has demonstrated the same level of FAR and FRR as that of Gunetti and Picardi approach while as high as 66.7% improvement of the authentication speed has been achieved.