Rumor Detection on Social Networks Based on the Degree Distribution Analysis in Step-by-Step Propagation Subgraphs

With the expansion of social networks and the increase in their users, these networks have become an effective medium for publishing news and various content. Therefore, new challenges have been created in this space, one of the most important of which is spreading rumors and false information. Rumors are moving at an incredible rate in society due to their appeal and attraction. Their spread can have many destructive effects on human societies and sometimes have irreparable consequences. For this reason, many researchers today deal with rumors in these networks. The purpose of this article is to provide a new method that can detect rumors without user information and post content analysis, and only according to the post propagation subgraph. Therefore, the degree distribution of the propagation graphs in the rumored and non-rumored models is examined. Then different classifiers were used to distinguish between these two modes. The Random Forest classifier gives better results than others. Since this method can finally detect rumors within four steps after propagation, this method has a good performance in terms of time.