Summary: | 碩士 === 國立中央大學 === 軟體工程研究所 === 98 === In recent years, with the popularization of computer and network, the single sign-on approach of the tradition authentication mechanism may not enough to protect our cyber assets in the future since the illegal cybercrimes access is still on the increase. Consequently, many researchers propose a re-authentication concept that uses the mouse dynamics to verify the user''s identity.
At the present time, mouse trajectories were collected from all applications or a specified application (such as Internet Explorer) for implementing verification models. However, the mouse trajectories include many types of activity and noise that will affect the verification accuracy. Therefore, this thesis proposes a method that only obtains the specific mouse trajectories to category trajectories and to filter noise for improving the verification accuracy of the implemented models. To prove the proposed method, this thesis conducts experiments to compare the verification rate from three types of trajectories collection. The experiment results show that the accuracy of the proposed method is better than the others. The successful verification rate in our model is 94%. This thesis technique can be applied to existing environmental and upgrade computer security.
|