Authentication And Keystroke Recognition Characteristic

As a biometric feature that does not require any special hardware to access, keystroke dynamics is a well known technique in security research. In order to secure web systems upon and after login, we did an in-depth research of the use of keystroke dynamics in user authentication and user identification on the Internet.We present a two-factor authentication scheme based on fixed-text keystroke dynamics in order to secure the widely used username/password authentication scheme. The new scheme seamlessly embeds keystroke dynamics collection and recognition process into the old scheme, upgrading existing authentication system to a two-factor authentication system while ensuring the same user experience. We theoretically analyzed the safety, feasibility and efficiency of the new scheme. Extensive experiments are conducted to justify the most critical part of the new scheme-keystrokes dynamics collection and recognition. The results proved that the new scheme is practically feasibility. This paper is the first to apply CRFs to the recognition of fixed-text keystroke dynamics. We made a comparison to the popular Mahalanobis distance based nearest neighbor model. The results showed that CRFs model outperforms the nearest neighbor model in recognition accuracy.We did a research on user identification based on free-typing keystroke dynamics. A mathematical detection model is defined upon the KS-test. The proposed model can identify existing users and distinguish between existing and new user. Experiments showed that our detection model was able to accurately identify both existing user and new user. The results confirmed that our detection model can accurately identify a user’s identity and distinguish between new and existing users. We also analyzed the effect of changing model parameters. Finally, we proposed several countermeasures to protect user privacy from free-typing keystroke dynamics analysis.