Reasearch On User Abnormal Behavior Detection And Node Threat Analysis In Online Social Platform

With the rapid development of the Internet in recent years,online social networks(OSNs)platforms have become an important part of people's daily life.OSNs have the characteristics of fast and instant.It combines with the interpersonal relationship of traditional social communication and shortens the distance between people.It is precisely because of such characteristics that many criminals can carry out attacks from the perspective of social engineering.This type of attack is highly targeted,threatening and difficult to defend,causing many losses to the platform and users.Therefore,the threat degree of the user node should be evaluated immediately,and the corresponding defense and processing should be performed.It is extremely important for users of ONSs to perform anomaly detection and judgment.From the perspective of social engineering,this paper sorts out the existing research results in related fields at home and abroad.Here,Twitter platform users are selected as the research dataset.The abnormal degree of user behavior is analyzed to evaluate whether the account is compromised or not.The calculation method of node influence is proposed,and then the threat degree of node is obtained.Firstly,based on the analysis and comparison of the existing research results,two categories of features of user behavior patterns are proposed.According to the actual needs of this study and the user's activities on the platform,7 features are proposed based on the message flow to construct the user profile.Secondly,an abnormal account compromised detection method based on Supervised Analytic Hierarchy Process(SAHP)is proposed.The specific abnormal score calculation method was proposed for the above 7 features.And the information gain ratio was used to rank the features,which provided quantitative support for the AHP,the weight of each feature was obtained.Then select different thresholds to detect the compromised of the account.Compared with the existing research results,the effectiveness of this method in evaluating anomaly detection is verified.Finally,the information transmission ability of the node is calculated to determine its influence in the social platform.The correlation between the experimental results and the actual situation is calculated to demonstrate the reliability of the experimental method.Combined with the above abnormal scores of nodes,the node threat degree is finally obtained.