Keystroke and Eye-Tracking Biometrics for User Identification

Two sources of weak biometric data are investigated for the development of user identification models. A probabilistic neural network model is created from keystroke digraph features extracted from raw typing data. A second model is developed from scan-path features extracted from raw eye-tracking data. A third model is created from a combination of features from the two biometric sources. Experimental results show that models based on keystroke biometric data perform very well, whereas the models involving the eye-tracking data are not as successful but encourage further study.