Centrality-Based Anomaly Detection on Multi-Layer Networks Using Many-Objective Optimization

Anomaly detection on complex network is receiving increasing attention, e. g., for finding illegal financial transactions, or for understanding the behavior of people via analyzing social network data. This paper presents a novel method for recognizing and finding anomalies in complex networks. Specifically, it targets multi-layer social network data aiming at finding abnormal behavior of some (groups of) nodes in the network. The method starts by measuring the centrality of all nodes in each layer of the multi-layer network, continues by applying many-objective optimization with full enumeration based on minimization, and obtains the Pareto Front. Objective functions to be optimized simultaneously are the centrality of each layer in the network and thus, the number of objective function are the numbers of existing layers of a multi-layer networks. After the Pareto Front settles, the set of nodes in the Pareto Front are considered as a basis for finding the set of suspected anomaly nodes, using the novel ACE-Score. The ACE-Score is calculated by considering the centrality of a node in the i - th layer, the mean of the centrality in that layer, the standard deviation, and the edge density of each layer. A high ACE-Score then indicates candidate anomalous nodes. We evaluate the approach on generated synthetic network as well as real-world complex networks, demonstrating the effectiveness of the proposed approach. A key feature of our proposed approach is its interpretability and explainability, since we can directly assess anomalous nodes with respect to the network topology.