A Graph-Based Approach to Detect Anomalies Based on Shared Attribute Values

Anomaly detection is an important task in many fields such as eHealth and online fraud. In this paper, we propose a new technique for anomaly detection based on a graph that connects transactions with the same attribute values and searches for dense clusters indicative of an anomalous pattern. The experimental evaluation shows that the graph-based approach outperforms two other approaches in the considered dataset. The extension of this approach to the eHealth domain is reserved as future work.