A method for abnormal user detection in social networks based on network mapping

Abstract

The invention relates to a method for detecting abnormal users in a social network based on network mapping, comprising the following steps: S1, using web crawler technology to crawl user data on a social network platform; S2, preprocessing the crawled user data, and constructing User social relationship network graph G; S3, based on the user social relationship network graph G, use node2vec to convert the user's social relationship into a low-dimensional vector representation; S4, integrate the user's multi-dimensional vector representation to obtain the final vector representation; S5, based on The user's feature vector is clustered to predict whether the user is a normal user or an abnormal user, and if it is an abnormal user, the abnormal type is given. The invention has the advantages of low time and labor costs, the ability to identify various abnormal user types and new abnormal user types, comprehensive consideration of multi-dimensional attribute characteristics of users, and high accuracy.

Description

technical field [0001] The invention relates to the technical field of network data detection, in particular to a method for detecting abnormal users in a social network based on network mapping. Background technique [0002] A series of excellent features such as convenience, entertainment, and real-time nature of social networks have attracted a large number of users, and a virtual society has been built in cyberspace. At the same time, the huge user base of social networks has attracted a large number of attackers. By creating a large number of fake accounts and stealing normal accounts, attackers spread rumors in social networks, or publish advertisements, phishing, pornographic information, etc., or use these accounts to maliciously increase the reputation of other accounts, such as mass follow, malicious points Like and so on. These attackers are collectively referred to as anomalous users. Therefore, abnormal user detection in social networks is of great significan...

AI Technical Summary

Problems solved by technology

Although these detection methods have been widely used in practice, these methods can only achieve good results in specific application scenarios, the generalization ability is not strong, and they are not universal, so they cannot meet the actual needs very well.

Specifically, although behavioral feature-based and content-based detection methods have high accuracy, they need to mark the sample data in advance because of the supervised learning method, which takes a lot of time and labor costs, and can only detect existing Known abnormal types, when the abnormal user changes its appearance, it cannot be accurately detected

Although the graph-based detection method has strong robustness, the accuracy rate is low, and it can only detect abnormal users who are connected with other users. It is still in the theoretical research stage.

The method based on unsupervised learning does not need to mark the sample data in advance, which saves time and labor costs, so it can quickly form a detection system and detect unknown attack behaviors, but it is not easy to distinguish different types of abnormal users

Images