Semi-supervised Anomaly Detection on Attributed Graphs

Abstract

We propose a simple yet effective method for detecting anomalous instances on an attribute graph with label information of a small number of instances. Although with standard anomaly detection methods it is usually assumed that instances are independent and identically distributed, in many real-world applications, instances are often explicitly connected with each other, resulting in so-called attributed graphs. The proposed method embeds nodes (instances) on the attributed graph in the latent space by taking into account their attributes as well as the graph structure based on graph convolutional networks (GCNs). To learn node embeddings specialized for anomaly detection, in which there is a class imbalance due to the rarity of anomalies, the parameters of a GCN are trained to minimize the volume of a hypersphere that encloses the node embeddings of normal instances while embedding anomalous ones outside the hypersphere. This enables us to detect anomalies by simply calculating the distances between the node embeddings and hypersphere center. The proposed method can effectively propagate label information on a small amount of nodes to unlabeled ones by taking into account the node's attributes, graph structure, and class imbalance. In experiments with five real-world attributed graph datasets, we demonstrate that the proposed method achieves better performance than various existing anomaly detection methods.

PhoGAD: Graph-based Anomaly Behavior Detection with Persistent Homology Optimization

“…These detection methods can be broadly classified into node-based detection and edge-based detection. [17] represents the most typical node-based detection method, involving the embedding of nodes into a lower-dimensional space, followed by a binary classifier for anomaly detection. Expanding upon this foundational framework, methods like CoLA [23] and LogLG [13] introduce novel indicators, such as the similarity between positive and negative sample pairs, to identify anomalies.…”

Graph Anomaly Detection With Graph Neural Networks: Current Status and Challenges

“…Semi-GCN [27] Label information → semi-supervised learning by GCN HCM [28] Label & contextual information → hop-count prediction model ResGCN [29] Over-smoothing issue → GCN with residual-based attention CoLA [30] Targeting issue of GAE → contrastive self-supervised learning ANEMONE [31] Contextual information → multi-scale contrastive learning PAMFUL [32] Contextual information → pattern mining algorithm with GCN GAT-based GAE AnomalyDAE [33] Complex interactions → GAT-based encoder GATAE [34] Over-smoothing issue → GAT-based encoder AEGIS [35] Handling unseen nodes → generative adversarial learning with GAE…”

“…Nonetheless, labeling information can help enhance the model perfor-mance. For this reason, Kumagai et al [27] proposed a simple yet effective GCN-based method, called semi-GCN, which is capable of embedding nodes into a hypersphere space while taking advantage of the structure and attribute features of a graph by stacking graph convolutional layers in order to detect global anomalies.…”

Graph Anomaly Detection with Graph Neural Networks: Current Status and Challenges

“…Nonetheless, labeling information can help enhance the model performance. For this reason, Kumagai et al [27] proposed a simple yet effective GCN-based method, called semi-GCN, which is capable of embedding nodes into a hypersphere space while taking advantage of the structure and attribute features of a graph by stacking graph convolutional layers in order to detect global anomalies.…”