Abnormal User Detection Based On Semi-Supervised Clustering

In recent years,with the rapid development of smart devices and the Internet,the Internet has increasingly penetrated into all aspects of people's lives,which has made the scale of users of various software applications gradually large.Nowadays,products with hundreds of millions of users are also common.At the same time,the value of users has gradually been valued,and the scale of active users has become one of the important indicators to measure the value of an Internet product.However,the huge user scale also attracts a large number of abnormal users who follow the benefits.The existence of abnormal users not only damages the security of normal users,but also seriously jeopardizes the reputation evaluation of software applications.Therefore,abnormal user detection is getting more and more attention.Common anomaly detection methods are mainly divided into three types:supervised methods,unsupervised methods,and semi-supervised methods.It is more difficult to tag a sufficient amount of data for anomalous users because their data is scarce.Therefore,although a supervised method can achieve better anomaly detection results,it requires a lot of manpower.Although the unsupervised method avoids this problem,the current unsupervised method fails to achieve the expected anomaly detection effect.Therefore,semi-supervised methods combined with supervision and unsupervised are gradually being applied to abnormal user detection.This paper proposes an abnormal user detection scheme based on semi-supervised clustering.The detection scheme combines the isolation forest algorithm and the semi-supervised fuzzy c-means clustering algorithm.First of all,cluster center is obtained from the labeled data,and the isolated data in the unlabeled data is separated,and then the possibility that the user belongs to a certain type is calculated according to the previously obtained cluster centers and isolated data,and finally semi-supervised clustering is performed using the obtained information as the priori information to detect the abnormal user.According to experiments,in the case of a small amount of labeled data,the detection scheme can also achieves a good detection effect.At the same time,according to this scheme,the user security system based on abnormal user detection is designed and implemented.The system includes modules such as data acquisition,abnormal user detection,system management and statistical analysis.The system test proves that the system can effectively detect abnormal users and run stably.