Research And Implementation Of User Identification Based On Mouse Dynamics

There are 3 methods of user authentication technologies:user name and password,ID cards or tokens,and biometric identification.User name and password is the most widely used authentication mechanisms because this authentication method can satisfy people's security needs without any extra cost,but this method can be easily stolen.ID cards or tokens also can achieve the purpose of certification,but it needs much more money to make a card,and ID cards and tokens may be lost.Biometric authentication is divided into two catalogs,one is the biological characteristics of the biological,the other is the biological behavior characteristics.Fingerprints and irises belong to biological characteristics of the biological,but they need some unique facilities and may have the potential to be deceived.Therefore,the trend of modern research has gradually shifted to the biological behavior characteristics.Biological behavioral characteristics include stride frequency,keystrokes,mouse dynamics and so on.Mouse dynamics stands out from other methods with its advantages of low cost,high user acceptance,no disclosure of sensitive information.The main problem in the existing user identification based on mouse dynamics is that it takes more time to collect user behavior data.In order to solve this problem,this paper uses a method of extracting mouse behavior characteristics based on time segment.Specifically,it uses a fixed length of time to slice user behaviors.Then,in each time segment,a statistical method is used to extract the user's mouse behavior characteristics.In addition,some new features like mouse left button click and right button click are utilized.According to the experiment,the distinction between the mouse left and right buttons can improve the accuracy of user authentication.Under Windows operating system,data of mouse behavior has been collected and analyzed,after that,combining linear discriminant analysis feature dimension reduction algorithm and support vector machine,a model has been built.This experiment collects 21 users' mouse behaviors data,for user identity identification,the experimental results show that mouse dynamics is effective for authenticating with a false accept rate(FAR)of 4.86%in 5 minutes,and the false rejection rate(FRR)is 5.56%.Besides,in this paper,an experiment proves the relationship between the type of pointing devices and the accuracy of user identification.