A Novel Graph Centrality Based Approach to Analyze Anomalous Nodes with Negative Behavior

Detection of different kinds of anomalous behaviors originating from negative ties among actors in online social networks is an unexplored area requiring extensive research. Due to increase in social crimes such as masquerading, bullying, etc., identification and analysis of these activities has become need of the hour. Approaches from two separate, yet, similar research areas, i.e. anomaly detection and negative tie analysis, can be clubbed together to identify negative anomalous nodes. Use of best measures from centrality based (negative ties) and structure based approaches (anomaly detection) can help us identify and analyze the negative ties more efficiently. A comparative analysis has been performed to detect the negative behaviors in online networks using different centrality measures and their relationship in curve fitting anomaly detection techniques. From results it is observed that curve fitting analysis of centrality measures relationship performs better than independent analysis of centrality measures for detecting negative anomalous nodes.