Feature-Based Anomaly Detection in Static Social Networks

The detection of anomalies in social networks has become a crucial data mining task because anomalies can provide valuable insights and actionable information for various purposes. This paper proposes a feature-based algorithm for detecting anomalous nodes in a community of a plain static social network. Our approach aims to identify anomalous nodes within a community by measuring the extent of their deviation from normal nodes. This proposed approach involves three main steps: first, detect the community in a network, second, compute an anomaly score for each node within the community based on its deviation from the expected behavior, and finally, identify anomalous nodes by comparing their anomaly scores to a predetermined statistical threshold. We also conducted an experiment using our algorithm to detect anomalous nodes within social network communities, which showed that calculating the anomaly score using two different centrality measures resulted in varying IDs for the anomalous nodes in the communities.