Anomaly Detection For User Behavior Based On WiFi Hotspots

Along the development of mobile communication network,users can use more and more ways to access to the network.WiFi and cellular network have become the main methods for users to access to the network.And,mobile Internet arise,research on user behavior analysis becomes the focus.In this paper,we call it a wireless reality analytics system,if the reality analytics system is realized through wireless technology.Also,we use this system based on WiFi technology to collect users' information to make anomaly analysis.Anomaly detection model in this project can be used in monitoring and directing of people overflow in public places,and detecting and conducting the passenger flow of shopping mallUsers' anomaly behavior means those behaviors which are in violation of history or typical users' behavior.Anomaly analysis of user behavior can be used in many fields.In the field of medicine,it can be used for remote medical care.In the field of public service,it can be used for theft behavior detection and emergency safety monitoring.In the field of traffic,it can be used for anomaly detection of vehicle overflow and direction.In the field of commerce,it can be used to make control of customer.For Anomaly Detection for User Behavior based on WiFi hotspots,the main work of this paper is described as follow:(1)Make summary of main research points of human mobility and comparison among WiFi,cellular and GPS technology to access user information.Make introduce of existing research on anomaly detection of user behavior.Then make analysis of common four detection methods in the fields of data dining.They are statistical model,distance model,relative density model and cluster method respectively.(2)Make summary of engineering basic of this research,the wireless reality analytics system.It has the characteristics of real-time,persistence,big data,and wide detection range and so on.This paper also introduces the three main modules and they are data collection module,data analysis module and Web presentation module.Based on existing problem,this paper provides optimization scheme and achievement method from two aspects of system architecture optimization and new function addition.(3)We use the cross entropy in Information theory to make anomaly detection for group users.There are some rules between time intervals.This paper makes research from the distribution of group user behaviors is considered in this paper.A novel method using cross entropy model which is usually used for networking attacking in computer network is proposed to detect anomalies of offline user behavior in wireless network.Moreover,to make the anomaly detection available in different scenes,cross entropy model is optimized by using membership function of fuzzy mathematic.Experiments conducted in campus,hotel and office satisfy the expected results,which verify the feasibility and efficiency.And the result retrieved from optimized model is better than that of original model.(4)For single user anomaly behavior,first,hierarchical clustering method is used to classify the users.Then the Bayesian discriminant formula is used to make joint analysis to identify four different types of users combining the current raw data.These data can influence users'dwell time in a period,the user's visiting frequency,and the user's signal.Moreover,different types of users in some of the targeted research is not normal,it can be used as an anomaly detection to achieve the purpose of data cleaning.(5)Some representative scenes of the application of user anomaly detection are descripted.