/Awesome-Graph-Anomaly-Detection

A collection of papers for graph anomaly detection, and published algorithms and datasets.

Awesome Graph Anomaly Detection

Collections for state-of-the-art (SOTA), novel awesome graph anomaly detecion methods (papers, codes and datasets)

We are looking forward for other participants to share their papers and codes. If interested, please contanct jingcan_duan@163.com or jinhu@nudt.edu.cn.

GitHub stars GitHub forks

Table of Contents

Important Survey and Benchmark Papers

  1. TKDE 2021: A Comprehensive Survey on Graph Anomaly Detection with Deep Learning [Paper]
  2. NeurIPS 2022: BOND: Benchmarking Unsupervised Outlier Node Detection on Static Attributed Graphs [Paper] [code]

Papers and Codes

Single-View Static Graph

Papers focus on node-level anomaly detection and work on single-view static graph datasets.

Traditional Methods

  1. SIGMOD 2000: LOF: Identifying Density-based Local Outliers [Paper] [code]
  2. KDD 2007: SCAN: a Structural Clustering Algorithm for Networks [Paper] [code]
  3. SDM 2016: Scalable Anomaly Ranking of Attributed Neighborhoods [Paper] [code]
  4. IJCAI 2017: Radar: Residual Analysis for Anomaly Detection in Attributed Networks [Paper] [code]
  5. IJCAI 2018: ANOMALOUS: A Joint Modeling Approach for Anomaly Detection on Attributed Networks [Paper] [code]

Deep Methods

Reconstruction
  1. SDM 2019: Deep Anomaly Detection on Attributed Networks [Paper] [code]
  2. DSAA 2021: ResGCN: Attention-based Deep Residual Modeling for Anomaly Detection on Attributed Networks [Paper] [code]
  3. TKDE 2021: Hybrid-order Anomaly Detection on Attributed Networks [Paper] [code]
  4. WSDM 2022: ComGA: Community-Aware Attributed Graph Anomaly Detection [Paper] [code]
Reinforcement Learning
  1. WSDM 2019: Interactive Anomaly Detection on Attributed Networks [Paper] [code]
  2. CIKM 2021: Towards Anomaly-resistant Graph Neural Networks via Reinforcement Learning [Paper]
Generative Adversarial Network
  1. IJCAI 2020: Inductive Anomaly Detection on Attributed Networks [Paper]
  2. CIKM 2020: Generative Adversarial Attributed Network Anomaly Detection [Paper]
Filter
  1. IJCAI 2022: Can Abnormality be Detected by Graph Neural Networks? [Paper] [code]
  2. ICML 2022: Rethinking Graph Neural Networks for Anomaly Detection [Paper] [code]
One-class SVM
  1. CIKM 2021: Subtractive Aggregation for Attributed Network Anomaly Detection [Paper]
  2. NCA 2021: One-Class Graph Neural Networks for Anomaly Detection in Attributed Networks [Paper] [code]
Meta Learning
  1. WWW 2021: Few-shot Network Anomaly Detection via Cross-network Meta-learning [Paper] [code]
Contrastive Learning
  1. TNNLS 2021: Anomaly Detection on Attributed Networks via Contrastive Self-Supervised Learning [Paper] [code]
  2. CIKM 2021: ANEMONE: Graph Anomaly Detection with Multi-Scale Contrastive Learning [Paper] [code]
  3. AAAI 2023: Graph Anomaly Detection via Multi-Scale Contrastive Learning Networks with Augmented View [Paper] [code]
Hybrid Methods
  1. TKDE 2021: Generative and Contrastive Self-Supervised Learning for Graph Anomaly Detection [Paper] [code]
  2. IJCAI 2022: Reconstruction Enhanced Multi-View Contrastive Learning for Anomaly Detection on Attributed Networks [Paper] [code]
  3. TKDE 2023: Counterfactual Graph Learning for Anomaly Detection on Attributed Networks [Paper]
Other Self-Supervised Learning
  1. Arxiv 2021: Hop-count Based Self-supervised Anomaly Detection on Attributed Networks [Paper] [code]
Others
  1. AAAI 2022: LUNAR: Unifying Local Outlier Detection Methods via Graph Neural Networks [Paper] [code]

Multi-View Static Graph

Papers focus on node-level anomaly detection and work on multi-view static graph datasets.

Reconstruction
  1. TKDE 2022: A Deep Multi-View Framework for Anomaly Detection on Attributed Networks [Paper]

Temporal Graph

Papers focus on node-level anomaly detection and work on single-view temporal graph datasets.

Others

Papers focus on graph-level anomaly detection and work on single-view static graph datasets.

  1. WSDM 2022: Deep Graph-level Anomaly Detection by Glocal Knowledge Distillation [Paper] [code]
  2. WSDM 2023: GOOD-D: On Unsupervised Graph Out-Of-Distribution Detection [Paper] [code]

Benchmark Datasets

Single-View Static Graph

Injected Datasets

By way of injection, adding anomalous nodes to datasets that do not have anomalies before. These anomalous nodes consist of feature anomalies and structure anomalies. The total number of anomalies are shown in the 5th column of table.

Dataset Nodes Edges Features Anomalies URL
BlogCatalog 5196 171743 8189 300 [BlogCatalog]
Flickr 7575 239738 12407 450 [Flickr]
ACM 16484 71980 8337 600 [ACM]
Cora 2708 5429 1433 150 [Cora]
Citeseer 3327 4732 3703 150 [Citeseer]
Pubmed 19717 44338 500 600 [Pubmed]

Real-world Anomaly Datasets

These datasets are born with anomalous nodes.

Datasets Nodes Edges Features Anomalies URL
Amazon 1418 3695 21 28 [Amazon]
Enron 13533 176987 20 5 [Enron]
YelpChi 45954 3846979 32 6677 [YelpChi]
T-Finance 39357 21222543 10 1803 [T-Finance]
T-Social 5781065 73105508 10 174010 [T-Social]
Elliptic 46564 73248 93 4,545 [Elliptic]

Other Related Awesome Repository

awesome-multi-view-clustering

Incomplete Multi-view Clustering

Awesome Deep Graph Clustering