Detecting User Interaction Anomaly based on Social Network Graph Similarity

Anomaly detection is one critical issue of analyzing the user interaction behaviors in social networks. This paper presents a method of detecting user interaction anomaly based on graph similarity. In the algorithm, we first establish a social network graph by using an existing social network dataset; then, according to the user connection relationship, the Louvain community discovery algorithm is used to divide the social network graph into multiple communities; next, we regard each community graph as an independent social network graph and calculate the quality score of each node and the feature value of each edge in the community graph; finally, we use the local hash to filter the feature value of each edge into the equal length binary feature number of each group of progressive comparisons. The community graphs with user abnormal interactions are obtained by comparing the thresholds according to the similarity. The experimental results show that our method can has good efficiency and accuracy of user interaction anomaly detection in the community.