The Research On Identity Recognition Schemes And Application Based On Multi-factor Behavior

Identity authentication and identification are collectively called identity recognition.Compared with traditional identity recognition,behavior-based identity recognition has gradually become a research hotspot due to its low cost and the fact that the recognition factors cannot be easily copied or stolen.However,the identity recognition of single factor behavior is vulnerable by other behaviors and the accuracy of identity recognition is low because of its single data characteristics.Firstly,in this paper,a recognition model based on multi-factor behavior is proposed to better recognize users by collecting multi-type behavior data.Secondly,a recognition method based on decision level fusion is proposed to improve the recognition performance.Finally,for most existing PC systems,once the user's identity is verified at the time of login,the system resources are always provided to the user,which is easy to be hijacked by attackers,making the security of the system threatened.Therefore,combining with the credibility model,a sustained recognition method based on the credibility model is proposed for the practical application of identity recognition,it dynamically recognize the user during the session after the user logs in.The specific research work is as follows:(1)This paper proposes a recognition model based on multi-factor behavior by combining keystroke,mouse and GUI.At the same time,the information divergence of features is used as the weight of the linear kernel function in support vector machines to obtain the I-SVM algorithm.The experiments show that the accuracy of multi-factor behavior identification is higher than single behavior,and the accuracy of I-SVM is higher than linear kernel function and gaussian kernel function of SVM.(2)This paper proposes a recognition method based on decision level fusion.In order to improve the performance of identity recognition,the idea of ensemblelearning Stacking was introduced and a two-layer classifier was used for recognize.The first-level classifier classifies the features of different behaviors respectively,combines the obtained classification prediction label and prediction probability as the input of the second-level classifier,and uses the second-level classifier to obtain the final prediction result.Compared with feature fusion,the experiment shows that the decision-level fusion method proposed in this paper is superior to feature fusion method in terms of identity recognition.(3)Based on the above research,combining with the trust model,a method of continuous identity recognition based on the credibility model is proposed,which enables users to conduct dynamic recognition of users in the background without feeling after login.FAR and FRR are mainly used to measure the performance of static identity authentication.For the performance of dynamic identity authentication,based on ANIA and ANGA,ATI and ATG are proposed in this paper to measure the performance of continuous identity authentication.Experiments show that ATG have good performance in dynamic identity authentication,the idea of continuous identity identification for multi-factor behavior is provided,it provides an idea for continuous recognition of multi-factor behaviors.