Research On Continuous Identity Authentication Based On Mouse And Window Behavior

As the foundation and important guarantee method of computer and network system security,user identity authentication has extensive research and application in practice.However,traditional authentication methods such as passwords are difficult to resist internal threats,so people have been trying to find a more efficient identity authentication method to meet the needs of system security.In recent years,the application of biometrics in authentication systems has received increasing attention.Biometrics are difficult to be imitated and cracked by people,and the security is extremely high.The technology is mainly divided into two categories,one is identity authentication based on physiological characteristics,but this type of identification technology requires specific hardware support,which will pay expensive costs.The other type is identity authentication based on user behavior,such as gait and mouse behavior.Among them,the identity authentication technology based on mouse behavior does not rely on additional hardware devices and can be directly deployed in many computer systems,so it is favored in the current security research field.Aiming at the problem of low authentication performance in the open environment of the current identity authentication method based on mouse dynamics,this paper proposes a solution for cross-domain analysis combining mouse and window behavior.The main research work and innovations are as follows:1)In order to make the identity authentication research based on mouse and window behavior more universally applicable,this paper builds a completely free experimental environment and designs a collector to collect user's mouse behavior data.The collection process is completely "transparent" to the user,without any restrictions on scenes and tasks,and the collected data can truly reflect the user's mouse dynamics.2)For the first time in this paper,the combination of window behavior features and mouse behavior features is used for continuous identity authentication.The combination of dualdomain features can better describe the user's behavior patterns in a short period of time.The experimental results show that,within the detection time of 2 minutes,the dual-domain feature achieves better authentication performance than the pure mouse dynamics feature,and the performance is superior to current research on identity authentication based on mouse behavior.3)Affected by subjective and objective factors,the user's mouse behavior is prone to change.This article defines the variability of mouse behavior,and proposes a voting feature selection algorithm to reduce this behavior variability.The final experimental results show that the voting feature selection algorithm can effectively improve the performance of the authentication system,and the feature space of the mouse and window behavior after feature selection is very stable.4)This paper analyzes the impact of different sample segmentation methods on classification performance.The experimental results show that sample partition based on data volume can make the authentication system show the best performance.