Anomaly Detection in Dynamic Graph based on Deep Graph Auto-encoder

Abstract

Dynamic networks are ubiquitous in daily life, such as the power data center and social network. Anomalies in dynamic networks seriously endanger the security of the network. Therefore, it is a critical task to detect anomalies in dynamic networks. This paper proposes an anomaly detection system based on network embedding learning, which encodes the dynamic network, learns the embedding vector of each node in the network, and performs anomaly detection by clustering the embedding vector. We propose a depth graph autoencoder model to learn the dynamic node embedding vectors. The we calculate the anomaly score based on the distance of the node to its nearest cluster center. Extensive experiments on real-life datasets are conducted to illustrate that proposed method outperforms state-of-the-art baselines. Compared with the existing methods, the method in this paper improves the AUC by up to 11%.