DeepAD: A Joint Embedding Approach for Anomaly Detection on Attributed Networks

Detecting anomalies in the attributed network is a vital task that is widely used, ranging from social media, finance to cybersecurity. Recently, network embedding has proven an important approach to learn low-dimensional representations of vertexes in networks. Most of the existing approaches only focus on topological information without embedding rich nodal information due to the lack of an effective mechanism to capture the interaction between two different information modalities. To solve this problem, in this paper, we propose a novel deep attributed network embedding framework named DeepAD to differentiate anomalies whose behaviors obviously deviate from the majority. DeepAD (i) simultaneously capture both of the highly non-linear topological structure and node attributes information based on the graph convolutional network (GCN) and (ii) preserve various interaction proximities between two different information modalities to make them complement each other towards a unified representation for anomaly detection. Extensive experiments on real-world attributed networks demonstrate the effectiveness of our proposed anomaly detection approach.