Research On User Authentication Technology Of Mobile Intelligent Terminal Based On Behavior Sequence

As mobile intelligent terminals are more and more widely used in people's daily work and learning,people have more requirements for their security.Especially,many authentication technologies,such as password,graphical password and fingerprint,are widely used to protect the mobile intelligent terminals.But as the password,graphics and fingerprint are one-time authentication,they can only ensure the credibility of the user when login,and can not guarantee the authenticity of the user's identity in the use,which leads to the risk of theft or impersonation in the use of the mobile intelligent terminal.User authentication based on behavior features is a new type of identity authentication technology,which can provide users with dynamic,sustainable and unaware authentication.It has become a hot research topic.However,at present,the authentication schemes proposed by researchers mainly take the user behavior independent as the premise and foundation of the analysis and authentication,and do not consider the inherent relevance between the actual behavior of the user.In this paper,a user authentication framework based on behavior sequence is proposed to solve the problem of identity authentication in the use of mobile intelligent terminal.Three aspects that user behavior sequence pattern mining,user behavior similarity measurement and user behavior pattern updating are studied in this paper.The main content and innovation of this paper are as follows:1.A MSLP-Tree based sequential pattern mining method under multiple support conditions is proposed.The traditional sequential pattern mining algorithm uses a single support threshold,neglects the difference between user behavior and affects the accuracy of user behavior analysis.A linear prefix tree structure MSLP-Tree with multiple support attributes is proposed,which can save all data information and facilitate subsequent mining;Based on depth first principle,we mine all frequent patterns and maximum patterns based on MSLP-Tree recursively.Experimental results show that MSLP-growth algorithm is much better than MS-GSP algorithm in mining efficiency and space consumption.2.A behavior similarity measure method based on behavior common subsequence similarity is proposed.Focus on the problem that the efficiency of the traditional comparison method is less efficient and less accurate,and most of the algorithms may not adapt to the matching of users' behavior sequence.the comparison efficiency is improved in the form of the matching of maximum user pattern and the original sequence;and the matching compensation and tolerance machine are introduced to reduce the comparison error;the best of behavior is combined.The overall similarity is obtained by combining the maximum behavior similarity and all behavior similarity to further improve the accuracy of similarity results.By calculating the similarity between the users' overall behavior,the uniqueness of the user is evaluated.Experimental results show that the recognition rate of user behavior can reach more than 95% by using sequence similarity algorithm,and the user uniqueness is high.3.A user behavior pattern updating method based on MSLP-Tree is proposed.Aiming at low efficiency of complete update and complex updating of user pattern,sequence updating is divided into three sub problems,item updating,support threshold updating and sequence updating by considering the change of user behavior,and the problem is greatly simplified.Taking the original result as the precondition of mining,only mining and updating the related parts,effectively reducing the unnecessary time overhead in the mining process,and ensuring the updated model accuracy.It is superior to complete updating in small-scale pattern updating.The experimental results show that the efficiency of SIUA_MT algorithm,SDUA_MT algorithm and IDUA_MT algorithm are better than the MSLP-growth algorithm,and the advantage is more obvious in the small-scale updating.