Mouse gesture based authentication using machine learning algorithm
G. Muthumari, R. Shenbagaraj, M. Blessa Binolin Pepsi · 2014
To authenticate a computer user, classification based on mouse operating behavior is proposed. It is used to perform the authentication task in initial login stage by capturing the mouse movements. The mouse operating behavior of users can be captured as coordinate axes and elapsed time, based on movement of the mouse. The obtained mouse behavior data consist of outliers and behavioral variability; these can be addressed by Peirce's criterion and Weighted Least Square Regression (WLSR) respectively. The features are extracted from that mouse behavior data to categorize the user's unique mouse behavior. The extracted features are analyzed and performs authentication using Learning Vector Quantization (LVQ). This LVQ Algorithm is used as a classifier, to find whether the given sample is authorized or unauthorized identity. The test result proves that, the proposed method WLSR with learning vector quantization classifier provides low error rates with good accuracy than the existing method.