Unveiling Anomalous Nodes Via Random Sampling and Consensus on Graphs

Abstract

The present paper develops a graph-based sampling and consensus (GraphSAC) approach to effectively detect anomalous nodes in large-scale graphs. GraphSAC randomly draws subsets of nodes, and relies on graph-aware criteria to judiciously filter out sets contaminated by anomalous nodes, before employing a semi-supervised learning (SSL) module to estimate nominal label distributions per node. These learned nominal distributions are minimally affected by the anomalous nodes, and hence can be directly adopted for anomaly detection. The per-draw complexity grows linearly with the number of edges, which implies efficient SSL, while draws can be run in parallel, thereby ensuring scalability to large graphs. GraphSAC is tested under different anomaly generation models based on random walks, as well as contemporary adversarial attacks for graph data. Experiments with real-world graphs showcase the advantage of GraphSAC relative to state-of-the-art alternatives.

BOND: Benchmarking Unsupervised Outlier Node Detection on Static Attributed Graphs

“…Besides the general node outlier definition, many previous works [2,3,17,29,42,47] have defined fine-grained node outlier types from different perspectives. In this paper, we unify the previous node outlier taxonomies as two major types: structural outlier and contextual outlier, which are illustrated in Figure 1 and defined below:…”

“…Graph outlier detection (OD) is a key machine learning (ML) task with many applications, including social network spammer detection [69], sensor fault detection [24], financial fraudster identification [19], and defense on graph adversarial attacks [29]. Different from classical outlier detection on tabular data and time-series data, graph outlier detection faces additional challenges: (1) complex data structure carries richer information, and thus more powerful ML models are needed to learn informative representations and (2) with more complex ML models, it often incurs higher training difficulty and more extensive memory consumption [31], posing challenges for time-critical (i.e., low time budget) and resource-sensitive (e.g., limited GPU memory) applications.…”