Skip to content

Latest commit

 

History

History
499 lines (316 loc) · 49.3 KB

README.md

File metadata and controls

499 lines (316 loc) · 49.3 KB
Raw

Deep Graph Anomaly Detection: A Survey and New Perspectives

Awesome PRs Welcome Stars Visits Badge GitHub forks

A professionally curated list of awesome resources (paper, code, data, etc.) on Deep Graph Anomaly Detection (DGAD), which is the first work to comprehensively and systematically summarize the recent advances of deep graph anomaly detection from the methodology design to the best of our knowledge.

We will continue to update this list with the latest resources. If you find any missed resources (paper/code) or errors, please feel free to open an issue or make a pull request.

Survey Paper

Deep Graph Anomaly Detection: A Survey and New Perspectives

Hezhe Qiao, Hanghang Tong, Bo An, Irwin King, Charu Aggarwal, Guansong Pang.

If you find this repository helpful for your work, please kindly cite our paper.

@article{qiao2024deep,
  title={Deep Graph Anomaly Detection: A Survey and New Perspectives},
  author={Qiao, Hezhe and Tong, Hanghang and An, Bo and King, Irwin and Aggarwal, Charu and Pang, Guansong},
  journal={arXiv preprint arXiv:2409.09957},
  year={2024}
}

Feel free to point out any mistakes and welcome to provide relevant papers.

Taxonomy of Deep Graph Anomaly Detection


Outline

The outline corresponds to the taxonomy of methods in our survey paper.

Categories of Deep Graph Anomaly Detection

1. GNN Backbone Design


1.1 Discriminative GNNs

1.1.1 Aggregation Mechanism

  • [Dou2020] Enhancing Graph Neural Network-based Fraud Detectors against Camouflaged Fraudsters in CIKM, 2020. [paper][code]

  • [Liu2020] Alleviating the Inconsistency Problem of Applying Graph Neural Network to Fraud Detection in SIGIR, 2020. [paper][code]

  • [Liu2021] Pick and Choose: A GNN-based Imbalanced Learning Approach for Fraud Detection in WWW, 2021.[paper][code]

  • [Zhang2021] FRAUDRE: Fraud Detection Dual-Resistant to Graph Inconsistency and Imbalance in ICDM, 2021. [paper][code]

  • [Zhang2022] Dual-discriminative Graph Neural Network for Imbalanced Graph-level Anomaly Detection in NeurIPS, 2022. [paper]

  • [Qin2022] Explainable Graph-based Fraud Detection via Neural Meta-graph Search in CIKM, 2022. [paper][code]

  • [Dong2022] Bi-Level Selection via Meta Gradient for Graph-based Fraud Detection in DASFAA, 2022. [paper]

  • [Shi2022] H2-FDetector: A GNN-based Fraud Detector with Homophilic and Heterophilic Connections in WebConf, 2022. [paper]

  • [Gao2023] Addressing Heterophily in Graph Anomaly Detection: A Perspective of Graph Spectrum in WebConf, 2023. [paper][code]

  • [Ma2023] Towards Graph-level Anomaly Detection via Deep Evolutionary Mapping in KDD, 2023. [paper][code]

  • [Chang2024] Multitask Active Learning for Graph Anomaly Detection in Arxiv, 2024. [paper][code]

  • [Zhang2024] Generation is better than Modification: Combating High Class Homophily Variance in Graph Anomaly Detection in Arxiv, 2024. [paper]

  • [Chen2024] Boosting Graph Anomaly Detection with Adaptive Message Passing in ICLR, 2024. [paper]

  • [Zhuo2023] Partitioning Message Passing for Graph Fraud Detection in ICLR, 2024. [paper][code]

  • [Gao2024] Graph Anomaly Detection with Bi-level Optimization in WebConf, 2024. [paper][code]

  • [Guo2024] Graph Local Homophily Network for Anomaly Detection in CIKM, 2024. [paper]

  • [Ma2024] Revisiting Score Propagation in Graph Out-of-Distribution Detection in NeurIPS, 2024. [paper][code]

  • [Lin2024] UniGAD: Unifying Multi-level Graph Anomaly Detection in NeurIPS, 2024. [paper][code]

1.1.2 Feature Transformation

  • [Chai2022] Can Abnormality be Detected by Graph Neural Networks? in IJCAI, 2022. [paper][code]

  • [Tang2022] Rethinking Graph Neural Networks for Anomaly Detection in ICML, 2022.[paper][code]

  • [Gao2023] Alleviating Structural Distribution Shift in Graph Anomaly Detection in WSDM, 2023.[paper][code]

  • [Dong2023] Rayleigh Quotient Graph Neural Networks for Graph-level Anomaly Detection in ICLR, 2024.[paper][code]

  • [Dong2024] SmoothGNN: Smoothing-based GNN for Unsupervised Node Anomaly Detection in Arxiv, 2024. [paper]

1.2 Generative GNNs

1.2.1 Feature Interpolation

  • [Zhao2021] GraphSMOTE: Imbalanced Node Classification on Graphs with Graph Neural Networks in WSDM, 2021 [paper][code]

  • [Han2022] G-Mixup: Graph Data Augmentation for Graph Classification in ICML, 2022 [paper][code]

  • [Park2022] GRAPHENS:Neighbor-aware Ego Network Synthesis for Class-imbalance Node Classification in ICLR, 2022. [paper][code]

  • [Liu2022] DAGAD: Data Augmentation for Graph Anomaly Detection in ICDM, 2022. [paper][code]

  • [Lou2023] GADY Unsupervised Anomaly Detection on Dynamic Graphs in Arxiv, 2023. [paper]

  • [Meng2023] Generative Graph Augmentation for Minority Class in Fraud Detection in CIKM, 2023. [paper]

  • [Zhou2023] Improving Generalizability of Graph Anomaly Detection Models via Data Augmentation in TKDE, 2023. [paper][code]

  • [Liu2024] Class-Imbalanced Graph Learning without Class Rebalancing in ICML, 2024. [paper][code]

  • [Chen2024] Consistency Training with Learnable Data Augmentation for Graph Anomaly Detection with Limited Supervision in ICLR, 2024. [paper][code]

  • [Zhou2024] Graph Anomaly Detection with Adaptive Node Mixup in CIKM, 2024. [paper]

  • [Kim2024] ANOMIX: A Simple yet Effective Hard Negative Generation via Mixing for Graph Anomaly Detection in Arxiv, 2024. [paper][code]

1.2.2 Noise Perturbation

  • [Cai2023] Self-Discriminative Modeling for Anomalous Graph Detection in Arxiv, 2023. [paper]

  • [Liu2023] GODM Data Augmentation for Supervised Graph Outlier Detection with Latent Diffusion Models in Arxiv, 2023. [paper][code]

  • [Lou2023] GADY: Unsupervised Anomaly Detection on Dynamic Graphs in Arxiv, 2023. [paper][code]

  • [Ma2024] Graph Anomaly Detection with Few Labels: A Data-Centric Approach in KDD, 2024. [paper]

  • [Qiao2024] Generative Semi-supervised Graph Anomaly Detection in NeurIPS, 2024. [paper][code]

  • [Li2024] DiffGAD: A Diffusion-based unsupervised graph anomaly detector in Arxiv, 2024. [paper][code]

2. Proxy Task Design


2.1 Graph Reconstruction

  • [Yu2018] NetWalk: A Flexible Deep Embedding Approach for Anomaly Detection in Dynamic Networks in KDD, 2018. [paper][code]

  • [Ding2019] Deep Anomaly Detection on Attributed Networks in SDM, 2019. [paper][code]

  • [Fan2020] ANOMALYDAE: Dual Autoencoder for Anomaly Detection on Attribute Networks in ICASSP, 2020. [paper][code]

  • [Bandyopadhyay2020] Outlier Resistant Unsupervised Deep Architectures for Attributed Network Embedding in WSDM, 2020. [paper][code]

  • [Pei2022] ResGCN Attention-based Deep Residual Modeling for Anomaly Detection on Attributed Networks in Machine Learning, 2022. [paper][code]

  • [Liu2022] Mul-GAD: a semi-supervised graph anomaly detection framework via aggregating multi-view information in Arxiv, 2022. [paper][code]

  • [Chen2022] AnomMAN: Detect Anomaly on Multi-view Attributed Networks in Information Sciences, 2022.[paper]

  • [Peng2022] A Deep Multi-View Framework for Anomaly Detection on Attributed Networks in TKDE, 2022. [paper]

  • [Luo2022] ComGA: Community-Aware Attributed Graph Anomaly Detection in WSDM, 2022. [paper][code]

  • [Zhang2022] Reconstruction Enhanced Multi-View Contrastive Learning for Anomaly Detection on Attributed Networks in IJCAI, 2022. [paper][code]

  • [Huang2022] Unsupervised Graph Outlier Detection: Problem Revisit, New Insight, and Superior Method in Arxiv, 2022. [paper][code]

  • [Niu2023] Graph-level Anomaly Detection via Hierarchical Memory Networks in ECML PKDD, 2023. [paper][code]

  • [Huang2023] Hybrid-Order Anomaly Detection on Attributed Networks in TKDE, 2023 [paper][code]

  • [Mesgaran2023] A graph encoder–decoder network for unsupervised anomaly detection in Arxiv, 2023. [paper]

  • [Kim2023] Label-based Graph Augmentation with Metapath for Graph Anomaly Detection in Arxiv, 2023. [paper][code]

  • [Roy2024] GAD-NR: Graph Anomaly Detection via Neighborhood Reconstruction in WSDM, 2024. [paper][code]

  • [He2024] ADA-GAD:Anomaly-Denoised Autoencoders for Graph Anomaly Detection in AAAI, 2024. [paper][code]

  • [Liu2024] STRIPE Spatial-temporal Memories Enhanced Graph Autoencoder for Anomaly Detection in Dynamic Graphs in Arxiv, 2024. [paper]

  • [Zou2024] A Structural Information Guided Hierarchical Reconstruction for Graph Anomaly Detection in CIKM, 2024. [paper]

  • [Kim2024] Rethinking Reconstruction-based Graph-Level Anomaly Detection: Limitations and a Simple Remedy in Arxiv, 2024. [paper]

2.2 Graph Contrastive Learning

  • [Jin2021] ANEMONE: Graph Anomaly Detection with Multi-Scale Contrastive Learning in CIKM, 2021. [paper]

  • [Zheng2021] Generative and Contrastive Self-Supervised Learning for Graph Anomaly Detection in TKDE, 2021. [paper][code]

  • [Liu2021] Anomaly Detection in Dynamic Graphs via Transformer in TKDE, 2021. [paper][code]

  • [Liu2021] CoLA Anomaly Detection on Attributed Networks via Contrastive Self-Supervised Learning in TNNLS, 2021. [paper][code]

  • [Xu2022] CONDA Contrastive Attributed Network Anomaly Detection with Data Augmentation in PAKDD, 2022. [paper][code]

  • [Wang2021] Decoupling Representation Learning and Classification for GNN-based Anomaly Detection in SIGIR, 2021. [paper][code]

  • [Chen2022] GCCAD:Graph Contrastive Coding for Anomaly Detection in TKDE, 2022. [paper][code]

  • [Wang2022] Cross-Domain Graph Anomaly Detection via Anomaly-aware Contrastive Alignment in AAAI, 2022. [paper][code]

  • [Zhang2022] Reconstruction Enhanced Multi-View Contrastive Learning for Anomaly Detection on Attributed Networks in IJCAI, 2022. [paper][code]

  • [Xu2023] Few-shot Message-Enhanced Contrastive Learning for Graph Anomaly Detection in Arxiv, 2023. [paper]

  • [Duan2023] ARISE: Graph Anomaly Detection on Attributed Networks via Substructure Awareness in TNNLS, 2023. [paper][code]

  • [Liu2023] BOURNE: Bootstrapped Self-supervised Learning Framework for Unified Graph Anomaly Detection in Arxiv, 2023. [paper]

  • [Ding2023] GOOD-D:On Unsupervised Graph Out-Of-Distribution Detection in WSDM, 2023. [paper][code]

  • [Duan2023] GRADATE:Graph Anomaly Detection via Multi-Scale Contrastive Learning Networks with Augmented View in AAAI, 2023. [paper][code]

  • [Singh2023] GraphFC:Customs Fraud Detection with Label Scarcity in Arxiv, 2023. [paper][code]

  • [Liu2023] Revisiting Graph Contrastive Learning for Anomaly Detection in Arxiv, 2023. [paper][code]

  • [Lin2023] Multi-representations Space Separation based Graph-level Anomaly-aware Detection in SSDBM, 2023. [paper][code]

  • [Liu2023] Towards Self-Interpretable Graph-Level Anomaly Detection in NeurIPS, 2023. [paper][code]

  • [Zhou2023] Learning Node Abnormality with Weak Supervision in CIKM, 2023. [paper]

  • [Kong2024] Federated Graph Anomaly Detection via Contrastive Self-Supervised Learning in TNNLS, 2024. [paper]

  • [Chen2024] Towards Cross-domain Few-shot Graph Anomaly Detection in ICDM, 2024. [paper]

  • [Niu2024] Zero-shot Generalist Graph Anomaly Detection with Unified Neighborhood Prompts in Arxiv, 2024. [paper][code]

  • [Cheng2024] Graph Pre-Training Models Are Strong Anomaly Detectors in Arxiv, 2024. [paper]

2.3 Graph Representation Distillation

  • [Ma2020] Deep Graph-level Anomaly Detection by Glocal Knowledge Distillation in CIKM, 2020. [paper][code]

  • [Lin2023] Discriminative Graph-level Anomaly Detection via Dual-students-teacher Model in Arxiv, 2023. [paper]

  • [Cai2024] FGAD: Self-boosted Knowledge Distillation for An Effective Federated Graph Anomaly Detection Framework in Arxiv, 2024. [paper]

2.4 Adversarial Graph Learning

  • [Chen2020] Generative Adversarial Attributed Network Anomaly Detection in CIKM, 2020. [paper][code]

  • [Ding2021] Inductive Anomaly Detection on Attributed Networks in IJCAI, 2021. [paper][code]

  • [Xiao2023] Counterfactual Graph Learning for Anomaly Detection on Attributed Networks in TKDE, 2023. [paper][code]

  • [Meng2023] Generative Graph Augmentation for Minority Class in Fraud Detection in Arxiv, 2023. [paper][code]

2.5 Score Prediction

  • [Pang2019] DevNet Deep Anomaly Detection with Deviation Networks in KDD, 2019. [paper][code]

  • [Ding2021] Few-shot Network Anomaly Detection via Cross-network in WebConf, 2021. [paper][code]

  • [Tian2023] SAD:Semi-Supervised Anomaly Detection on Dynamic Graphs in IJCAI, 2023. [paper][code]

  • [Zhou2023] Learning Node Abnormality with Weak Supervision in CIKM, 2023. [paper]

  • [Xu2024] MetaGAD: Meta Representation Adaptation for Few-Shot Graph Anomaly Detection in DSAA, 2024. [paper][code]

3. Graph Anomaly Measures


3.1 One-class Classification Measure

  • [Teng2018] Deep into Hypersphere: Robust and Unsupervised Anomaly Discovery in Dynamic Networks in IJCAI, 2018. [paper][code]

  • [Wang2021] One-Class Graph Neural Networks for Anomaly Detection in Attributed Networks in * Neural Comput & Applic*, 2021. [paper][code]

  • [Zhou2021] Subtractive Aggregation for Attributed Network Anomaly Detection in CIKM, 2021. [paper][code]

  • [Li2023] HRGCN: Heterogeneous Graph-level Anomaly Detection with Hierarchical Relation-augmented Graph Neural Networks in DSAA, 2023. [paper][code]

  • [Zhang2023] Deep Graph-level Orthogonal Hypersphere Compression for Anomaly Detection in ICLR, 2024. [paper][code]

3.2 Community Adherence

  • [Yu2018] NetWalk: A Flexible Deep Embedding Approach for Anomaly Detection in Dynamic Networks in KDD, 2018. [paper][code]

  • [Zhou2022] Unseen Anomaly Detection on Networks via Multi-Hypersphere Learning in SDM, 2022. [paper][code]

3.3 Local Affinity

  • [Kim2023] Class Label-aware Graph Anomaly Detection in CIKM, 2023. [paper][code]

  • [Pan2023] PREM: A Simple Yet Effective Approach for Node-Level Graph Anomaly Detection in ICDM, 2023. [paper][code]

  • [Qiao2023] Truncated Affinity Maximization: One-class Homophily Modeling for Graph Anomaly Detection in NeurIPS, 2023. [paper][code]

  • [Liu2024] ARC: A Generalist Graph Anomaly Detector with In-Context Learning in Arxiv, 2024. [paper]

  • [Wang2024] Context Correlation Discrepancy Analysis for Graph Anomaly Detection in TKDE, 2024. [paper]

3.4 Graph Isolation

  • [Xu2023] Deep Isolation Forest for Anomaly Detection in TKDE, 2023. [paper][code]

  • [Zhuang2023] Subgraph Centralization: A Necessary Step for Graph Anomaly Detection in SDM,2023. [paper][code]

4. Related Surveys on Graph Anomaly Detection

  • [Ma2021] A Comprehensive Survey on Graph Anomaly Detection with Deep Learning in TKDE, 2021. [paper]

  • [Liu2022] BOND: Benchmarking Unsupervised Outlier Node Detection on Static Attributed Graphs in NeurIPS, 2022. [paper][code]

  • [Tang2023] GADBench: Revisiting and Benchmarking Supervised Graph Anomaly Detection in NeurIPS, 2023. [paper][code]

  • [Liu2023] A survey of imbalanced learning on graphs: Problems, techniques, and future direction in Arxiv, 2024. [paper]

  • [Ekle2024] Anomaly Detection in Dynamic Graphs: A Comprehensive Survey in Arxiv, 2024. [paper]

  • [Wang2024] Unifying Unsupervised Graph-Level Anomaly Detection and Out-of-Distribution Detection:A Benchmark in Arxiv, 2024. [paper][code]

5. Related Surveys on Anomaly Detection

  • [Pang2021] Deep Learning for Anomaly Detection: A Review in CSUR, 2021. [paper]

  • [Jiang2023] Weakly Supervised Anomaly Detection: A Survey in Arxiv, 2023. [paper][code]

  • [Xu2024] Large Language Models for Anomaly and Out-of-Distribution Detection: A Survey in Arxiv, 2024. [paper]

6. Quantitative Comparison

Quantitative comparison of node-level anomaly detection on datasets with manually injected (synthetic) anomalies

Metric Setting Cora Citeseer ACM BlogCatalog Flicker Pubmed Facebook Reddit Weibo
AUROC DOMINANT [Ding2019] 0.815 0.825 0.760 0.746 0.744 0.808 0.554 0.560 0.850
AUROC CoLA [Liu2021] 0.878 0.896 0.823 0.785 0.751 0.951 / 0.603 /
AUROC SL-GAD [Zheng2021] 0.913 0.913 0.853 0.818 0.796 0.967 / 0.567 /
AUROC CONAD [Xu2022] 0.788 / / / / / 0.863 0.561 0.854
AUROC AEGIS [Ding2021] / / / 0.743 0.738 0.773 / / /
AUROC OCGNN [Wang2021] 0.881 0.856 / / / 0.747 0.793 / /
AUROC ComGA [Luo2022] 0.884 0.916 0.849 0.814 0.799 0.922 0.659 / /
AUROC AAGNN [Zhou2021] / / / 0.818 0.829 0.856 / / 0.925
AUROC HCM-A [Huang2022] / / 0.761 0.798 0.792 / / / /
AUROC GAAN [Chen2020] 0.742 / 0.877 0.765 0.753 / / 0.554 0.925
AUROC AnomalyDAE [Fan2020] 0.762 0.727 0.778 0.783 0.751 0.810 / 0.557 0.915
AUROC GAD-NR [Roy2023] 0.835 / / / / / / / 0.623
AUROC TAM [Qiao2023] / / 0.887 0.824 / / 0.914 0.602 /
AURPC DOMINANT [Ding2019] 0.200 / / 0.338 0.324 0.299 / 0.037 /
AURPC CoLA [Liu2021] / / 0.323 0.327 / / 0.211 0.044 /
AURPC SL-GAD [Zheng2021] / / / 0.388 0.378 / 0.131 0.041 /
AURPC CONAD [Xu2022] / / / / / / / 0.037 /
AURPC AEGIS [Ding2021] / / / 0.339 0.324 0.373 / / /
AURPC OCGNN [Wang2021] / / / / / / / / /
AURPC ComGA [Luo2022] / / / / / / / / /
AURPC AAGNN [Zhou2021] / / / 0.435 0.421 0.428 / / /
AURPC HCM-A [Huang2022] / / / / / / / / /
AURPC GAAN [Chen2020] / / / 0.338 0.324 0.337 / 0.037 /
AURPC AnomalyDAE [Fan2020] 0.183 / / / / / / / /
AURPC GAD-NR [Roy2023] / / / / / / / / /
AURPC TAM [Qiao2023] / / 0.512 0.418 / / 0.223 0.044 /

Quantitative comparison of node-level anomaly detection on datasets with genuine anomalies

Metric Setting Amazon YelpChi T-Finance Question Elliptic Reddit Tolokers Weibo DGraph T-Social Photo CS
AUROC Unsupervised DOMINANT [Ding2019] 0.694 0.539 0.538 / 0.296 0.556 / / 0.574 / 0.514
AUROC Unsupervised CoLA [Liu2021] 0.261 0.480 0.483 / / 0.603 / / / / /
AUROC Unsupervised CLAD [Kim2023] 0.203 0.476 0.139 0.621 0.419 0.578 0.406 / / / /
AUROC Unsupervised GRADATE [Duan2023] 0.478 0.492 0.406 0.554 / 0.526 0.537 / / / /
AUROC Unsupervised GAD-NR [Roy2023] 0.260 0.470 0.579 0.587 0.400 0.553 0.576 / / / /
AUROC Unsupervised Prem [Pan2023] 0.278 0.490 0.448 0.603 0.497 0.551 0.565 / / / /
AUROC Unsupervised TAM [Qiao2023] 0.802 0.548 0.690 0.504 / 0.572 0.469 / / / /
AUROC Unsupervised SmoothGNN [Dong2024] 0.840 0.575 0.755 0.644 0.572 0.594 0.687 / 0.649 0.703 /
AUROC Semi-supervised GGAD [Qiao2024] 0.944 / 0.823 / 0.729 / / / 0.594 / 0.648
AUROC Supervised BWGNN [Tang2022] 0.980 0.849 0.961 0.718 0.852 0.654 0.804 0.973 0.763 0.920 /
AUROC Supervised DCI [Wang2021] 0.946 0.778 0.868 0.692 0.828 0.665 0.755 0.942 0.747 0.808 /
AUROC Supervised AMNet [Chai2022] 0.970 0.826 0.937 0.681 0.773 0.684 0.768 0.953 0.731 0.536 /
AUROC Supervised GHRN [Gao2023] 0.981 0.853 0.96 0.718 0.854 0.660 0.804 0.967 0.761 0.790 /
AUROC Supervised NGS [Qin2022] 0.973 0.921 / / / / / / / / /
AUROC Supervised PCGNN [Liu2021] 0.973 0.797 0.933 0.699 0.858 0.532 0.728 0.902 0.720 0.692 /
AUROC Supervised GDN [Gao2023] 0.971 0.903 / / / / / / / / /
AUROC Supervised DevNet [Pang2019] / / 0.654 / / / / / / / 0.599
AUROC Supervised PReNet [Pang2023] / / 0.892 / / / / / / / 0.698
AUROC Supervised NSReg [Wang2023] / / 0.929 / / / / / / / 0.908
AUPRC Unsupervised DOMINANT [Ding2019] 0.102 0.165 0.047 / / 0.036 / 0.008 / 0.104
AUPRC Unsupervised CoLA [Liu2021] 0.052 0.136 0.041 / / 0.045 / / / / 0.246
AUPRC Unsupervised CLAD [Kim2023] 0.040 0.128 0.025 0.051 0.081 0.050 0.192 / / / /
AUPRC Unsupervised GRADATE [Duan2023] 0.063 0.145 0.038 0.035 / 0.039 0.236 / / / /
AUPRC Unsupervised GADNR [Roy2023] 0.042 0.139 0.054 0.057 0.077 0.037 0.299 / / / /
AUPRC Unsupervised Prem [Pan2023] 0.074 0.137 0.039 0.043 0.090 0.041 0.259 / / / /
AUPRC Unsupervised TAM [Qiao2023] 0.332 0.173 0.128 0.039 / 0.042 0.196 / / / /
AUPRC Unsupervised SmoothGNN [Dong2024] 0.395 0.182 0.140 0.059 0.116 0.043 0.351 / 0.019 0.063 /
AUPRC Semi-supervised GGAD [Qiao2024] 0.792 / 0.183 / 0.243 0.061 / / 0.008 / 0.144
AUPRC Supervised BWGNN [Tang2022] 0.891 0.551 0.866 0.167 0.260 0.069 0.497 0.930 0.040 0.549 /
AUPRC Supervised DCI [Wang2021] 0.815 0.395 0.626 0.141 0.254 0.061 0.399 0.896 0.036 0.138 /
AUPRC Supervised AMNet [Chai2022] 0.873 0.488 0.743 0.146 0.147 0.073 0.432 0.897 0.028 0.031 /
AUPRC Supervised GHRN [Gao2023] 0.895 0.566 0.866 0.167 0.277 0.072 0.499 0.918 0.04 0.163 /
AUPRC Supervised NGS [Qin2022] / / / / / / / / / / /
AUPRC Supervised PCGNN [Liu2021] 0.878 0.437 0.698 0.144 0.356 0.042 0.381 0.819 0.028 0.087 /
AUPRC Supervised DevNet [Pang2019] / / 0.323 / / / / / / / 0.468
AUPRC Supervised PReNet [Pang2023]} / / 0.571 / / / / / / / 0.460
AUPRC Supervised NSReg [Wang2023] / / 0.757 / / / / / / / 0.836

Quantitative comparison of graph-level anomaly detection

Metric Methods PROTEINS-F ENZYMES AIDS DHFR BZR COX2 DD NCI1 IMDB COLLAB HSE MMP P53 TraceLog FlowGraph
AUROC GlocalKD [Ma2022] 0.773 0.613 0.932 0.567 0.694 0.593 0.801 0.684 0.521 0.674 0.593 0.675 0.640 / /
AUROC OCGIN [Zhao2023] 0.708 0.587 0.781 0.492 0.659 0.535 0.722 0.719 0.601 / / / / / /
AUROC SIGNET [Liu2024] 0.752 0.629 0.972 0.740 0.814 0.714 0.727 0.748 0.664 / / / / / /
AUROC OCGTL [Qiu2022] 0.765 0.620 0.994 0.599 0.639 0.552 0.794 0.734 0.640 / / / / / /
AUROC OCGCN [Wang2021] 0.718 0.613 0.664 0.495 0.658 0.628 0.605 0.627 0.536 / 0.388 0.457 0.483 / /
AUROC HimNet [Niu2023] 0.772 0.589 0.997 0.701 0.703 0.637 0.806 0.686 0.553 0.683 0.613 0.703 0.646 / /
AUROC GLADST [Lin2023] / 0.694 0.976 0.773 0.810 0.630 / 0.681 / 0.776 0.547 0.685 0.688 / /
AUROC DIF [Xu2023] / / / / / / / / / / 0.737 0.715 0.680 / /
AUROC HRGCN [Li2023] / / / / / / / / / / / / / 0.864 1.000

7. Datasets

Dataset # Nodes # Edges # Attributes Size Anomaly Anomaly Type Domain Download Link
Cora 2,708 5,429 1,433 Small 5.5% Injected Citation Networks [Link]
Citersee 3,327 4,732 3,703 Small 4.5% Injected Citation Networks [Link]
ACM 16,484 71,980 8,337 Medium 3.6% Injected Citation Networks [Link]
BlogCatalog 5,196 171,743 8,189 Small 5.8% Injected Social Networks [Link]
Flickr 7,575 239,738 12,407 Medium 5.2% Injected Social Networks [Link]
OGB-arXiv 169,343 1,166,243 128 Large 3.5% Injected Citation Networks [Link]
Amazon 11,944 4,398,392 25 Large 9.5% Genuine Transaction Record [Link]
YelpChi 45,954 3,846,979 32 Large 14.5% Genuine Reviewer Interaction [Link]
T-Finance 39,357 21,222,543 10 Large 4.6% Genuine Transaction Record [Link]
T-Social 5,781,065 73,105,508 10 Large 3.0% Genuine Social Network [Link]
Weibo 8,405 407,963 400 Small 10.3% Genuine Under Same Hashtag [Link]
DGraph 3,700,550 4,300,999 17 Large 1.3% Genuine Loan Guarantor [Link]
Elliptic 203,769 234,355 166 Large 9.8% Genuine Payment Flow [Link]
Tolokers 11,758 519,000 10 Medium 21.8% Genuine Work Collaboration [Link]
Questions 48,921 153,540 301 Medium 3.0% Genuine Question Answering [Link]
Disney 124 335 28 Small 4.8% Genuine Co-purchase [Link]
Books 1,418 3,695 21 Small 2.0% Genuine Co-purchase [Link]
Enron 13,533 176,987 18 Medium 0.4% Genuine Email network [Link]
Reddit 10,984 168,016 64 Medium 3.3% Genuine User-subreddit [Link]
Dataset # Graphs # Avg. Nodes # Edges Anomaly Domain Homo./Heter. Download Link
KKI 83 190 237.4 44.6% Bioinformatics Homo. [Link]
OHSU 79 82.01 199.66 44.3% Bioinformatics Homo. [Link]
MUTAG 188 17.93 19.79 33.5% Molecules Homo. [Link]
PROTEINSfull 1,113 39.06 72.82 40.4% Bioinformatics Homo. [Link]
ENZYMES 600 32.63 62.14 16.7% Bioinformatics Homo. [Link]
AIDS 2,000 15.69 16.2 20.0% Chemical Structure Homo. [Link]
BZR 405 35.75 38.36 21.0% Molecules Homo. [Link]
COX2 467 41.22 43.45 21.8% Molecules Homo. [Link]
DD 1,178 284.32 715.66 41.3% Bioinformatics Homo. [Link]
NCI1 4,110 29.87 32.3 49.9% Molecules Homo. [Link]
IMDB 1,000 19.77 96.53 50.0% Social Networks Homo. [Link]
REDDIT 2,000 429.63 497.75 50.0% Social Networks Homo. [Link]
HSE 8,417 16.89 17.23 5.2% Molecules Homo. [Link]
MMP 7,558 17.62 17.98 15.6% Molecules Homo. [Link]
p53 8,903 17.92 18.34 6.3% Molecules Homo. [Link]
PPAR-gamma 8,451 17.38 17.72 2.8% Molecules Homo. [Link]
COLLAB 5,000 74.49 2,457.78 15.5% Social Networks Homo. [Link]
Mutagenicit 4,337 30.32 30.77 44.6% Molecules Homo. [Link]
DHFR 756 42.43 44.54 39.0% Molecules Homo. [Link]
TraceLog 132,485 205 224 17.6% Log Sequences Heter. [Link]
FlowGraph 600 8,411 12,730 16.7% System Flow Heter. [Link]