Research On The Algorithm Of Identifying Social Network Sybil Attack Based On Data Mining

With the rapid development of online social networking,the online social network has already become an essential way of life for most people.Driven by interests,attacks on social networks are becoming more and more frequent.Compared with a single user,Sybil attacks launched by multiple users in collusion are more destructive.Sybil user group is characterized by strong organization and structure.Users in the group will launch Sybil attacks on the same target store almost at the same time.This Sybil attack can cause great changes in the reputation score of the target store in a short period of time.The sharp rise or fall in reputation score depends on whether the store itself or its competitor is the employer of Sybil activities.No matter what the result is,it has done harm to the normal market order.Therefore,there is a need for an effective method to identify Sybil attacks in online social networks.This paper mainly studies how to effectively identify Sybil attacks in social networks.In this paper,a scheme based on improved logistic regression algorithm to identify Sybil attacks on social networking sites is designed,taking the public comment network,a social networking site for user evaluation,as the research object.Firstly,the user reviews data set was obtained by crawling the user's comment data on top-ranked merchants on Dian Ping Online for a period of time.The collection of data uses the existing online collection tools.Secondly,the acquired data set is cleaned,including the dirty data such as missing values,abnormal values,duplicate data and so on.The evaluation data set can be used to prepare for the following work such as feature analysis,model construction and experimental verification.Then the data characteristics are analyzed and studied,mainly the user attribute characteristics and user behavior characteristics,including the user level,the overall rating of the store and the rating of service environment,whether the user is a member,the evaluation time,the number of comments received from others and the number of replies received from others,etc.The structure of these features is analyzed and processed,and the data distribution of each feature is studied.The Sybil attack recognition model can be effectively constructed by adding analysis of user behavior attributes.Identifying Sybil attacks on social networking sites can be simplified to classify all users and identify users as normal users or Sybil users.However,logistic regression algorithm is very suitable for classification,so a model based on logistic regression algorithm to identify Sybil attacks onsocial networking sites is constructed.In order to prevent over-fitting,the algorithm is improved by adding a regular term.The processed data set is divided into training set and test set,and the constructed model is trained.Experiments show that the improved logistic regression algorithm can better identify Sybil users.