Research On Identity Authentication Method Based On User Mouse Behavior

With the rapid development of Internet technology,security issues have attracted the attention of major companies and government agencies around the world.Identity authentication is an important guarantee for information system security.However,traditional identity authentication methods have the disadvantages of easy disclosure and easy loss of authentication factors.Therefore,identity authentication based on user biometrics has gradually become a hot topic in the field of identity authentication research in recent years.This thesis uses a multi-classifier fusion method to perform dynamic continuous identity authen tication tasks based on the user's mouse behavior characteristics and obtain a better authenticati on results.This article uses the background client software to collect the user's daily mouse behavior data and conduct experiments.The data collection and dynamic continuous identity authenticati on process are performed in the environment where the user's mouse operation is completely un controlled.The original mouse behavior data is ensured by the data preprocessing such as data c leaning and data transformation to ensure the integrity and rationality of the mouse behavior dat a.In view of the lack of mouse behavior feature space,this paper analyzes the user's mouse beh avior from two aspects:mouse's overall behavior characteristics and mouse's trajectory behavio r characteristics.Based on the analysis of mouse behavior,the original mouse behavior feature s pace is constructed.In order to reduce the influence of the user's mouse behavior variability on t he identity authentication result,a feature selection method is applied to obtain a more compact and stable mouse behavior feature space.In order to improve the performance of identity authe ntication and avoid the problems of overfitting and inaccurate classification accuracy of a single classifier,this thesis uses a multi-classifier fusion method to perform dynamic continuous identi ty authentication tasks based on the user's mouse behavior characteristics and obtain a better aut hentication results.In order to verify the effectiveness of the mouse behavior feature set constru cted in this thesis and the efficiency of the identity authentication method proposed in this thesi s,multi-angle comparison experiments are performed based on different data sets.The evaluati on results of the commonly used evaluation indicators FAR and FFR in the field of identity auth entication and the new evaluation indicators ANIA and ANGA presented in this thesis show that the mouse behavior characteristics proposed in this thesis combined with the identity authentic ation method of this study can complete user identity authentication with high accuracy.Based on the method of this thesis,we combine the dynamic trust model and identity authentication,and use the "reward and punishment" strategy to improve the fault tolerance and applicability of this method in practical applications.Simulating the process of dynamic identity authentication in actual application scenarios,the experiment proves that the accuracy rate of this method reaches more than 90%,and it can maintain high robustness to different application scenarios.