Social media abnormal group user detection method based on relation evolution

Abstract

The invention discloses a social media abnormal group user detection method based on relation evolution, and the method comprises the following steps: 1, carrying out m times of edge sampling on eachimage object in an image stream, and generating a stream sample each time of sampling; 2, constructing m node division modes for each graph object according to the flow sample; 3, constructing an edgeprobability model according to node division, and calculating composite edge likelihood fitting of each edge in the total sample edge set; and 4, calculating the likelihood fitting of each graph object according to the composite edge likelihood fitting, wherein the graph object with larger likelihood fitting is regarded as an abnormal graph. Aiming at a scene where group users have abnormal behaviors in social media, the improved abnormal group user detection method based on the weighted graph is provided, and the improved method can be used for processing a social media user interaction state network based on the weighted graph, so that the effectiveness of edge collection can be ensured, and abnormal false alarms caused by unintentional interaction of users can be reduced.

Description

technical field [0001] The invention belongs to the security field of social media abnormal user detection, and relates to a group abnormal user detection method based on relationship evolution in social media. Background technique [0002] In recent years, a large number of social applications have emerged and developed rapidly. For example, domestic well-known ones include Tencent QQ, WeChat, Sina Weibo, Baidu Tieba, Douban, Tianya Community, Zhihu, etc. Foreign well-known professional social networking sites LinkedIn, Weibo, etc. Social networking site Twitter, light blogging social platform Tumblr, the world's largest social networking site Facebook, image-based social networking site Pinterest, SNS social networking site Google+, etc. These social applications enable users to interact easily no matter where they are, and enable strangers who have never met before to find friends and confidants with similar interests. It can increase the communication frequency of friend...

AI Technical Summary

Problems solved by technology

[0005] The closeness of the relationship between users in social media can be reflected in the frequency of user interaction, while the non-weighted graph cannot consider the number of user interactions, so it is impossible to judge the closeness of the interaction relationship between users

Taking Weibo as an example, most ordinary users are accustomed to like and forward related content of star users. Although the number of interactive relationships generated by such interactive behaviors is huge, the relationship between these two types of users often not tight enough

Because the ODBGSCM method cannot identify the closeness of user relationship, it will collect a large number of edges with low user interaction relationship, and these edges cannot reflect the real structural characteristics of the graph, so it is meaningless for anomaly detection, and it will also reduce the algorithm processing efficiency

[0006] (2) There are abnormal false positives

[0007] Since the ODBGSCM method regards a graph containing a large number of edges across dense partitions as a suspicious abnormal graph object, as long as users in different dense partitions interact with each other, the graph may be reported as abnormal. However, such interactions may be There are few unintentional operations by users, such as unintentional likes, attention, etc., so there are false positives

Images