Identity Authentication based on User Mouse Behavior

Improving the performance of identity authentication is a difficult issue due to the sparsity of feature space of user mouse behaviors, this paper investigates a new identity authentication method, which improves the sparse representation of user mouse behaviors from a new perspective. Firstly, the feature space of user mouse behaviors can be characterized from both the mouse's overall behavior features and mouse’s trajectory behavior features and also can be effectively reduced according to the validity of features. Secondly, we present a fusion method of multiple classifiers to solve the problem of over-fitting in a single classifier and improve the performance of mouse-behavior based continuous authentication. Finally, comparative experiments are conducted in order to further examine the effectiveness of the proposed method. These findings suggest that the fusion method of multiple classifiers based on Stacking outperforms the traditional single classification method in user mouse identity identification.