The Study And Implement On Abnormal User Detection Of Online Social Network Based On KMV Model

Abnormal user will exist permanently in Online Social Network. In modern society, abnormal user detection is based on feature characteristics of information itself. Using methods those interface based, rank based or data feature based to detect abnormal users. But methods interface based and rank based need user extra active operation, will affect the use of online social network, data feature based aimed at data mining of historical data. KMV model is one of widely used credit monitor models, KMV based on the historical data and realtime data of subject, using Black-Scholes to get Expected Default Frequency(EDF) of the subject.This paper using KMV to abnormal user detection of online social network to:(1)Using credit monitor models to measure trust rank of users of online social network, make it more convenience for user classification based on trust rank in online social network.(2)Combined historical data with realtime data, benefit from low cost of KMV to detect abnormal actions of user in realtime.Main contribution of this paper includes the following aspects:First, we studied the development and characteristics of online social network, analysis user characteristics and action statistics of some of primary online social networks.Secondly, we studied the harm of abnormal user and solutions of abnormal user detection in online social network, analysis the difference of different abnormal user detection.Then, we studied the development of credit monitor method and model, focus on the parameter selection and definition of parameters, prepared to mapping KMV to online social network.Finally, we mapping KMV to abnormal user detection in online social network, analysis the result and proposed a new realtime abnormal user detection in online social network with lower computing work and maintaining the measurement accuracy.