A Zero-shot Generalized Graph Anomaly Detection Framework via Node Reconstruction

Dat Cao; Hien Chu; Khue Hoang; Phan Nguyen

arxiv: 2606.12673 · v1 · pith:ZXYQRXUJnew · submitted 2026-06-10 · 💻 cs.LG · cs.AI

A Zero-shot Generalized Graph Anomaly Detection Framework via Node Reconstruction

Phan Nguyen , Dat Cao , Hien Chu , Khue Hoang This is my paper

Pith reviewed 2026-06-27 10:10 UTC · model grok-4.3

classification 💻 cs.LG cs.AI

keywords graph anomaly detectionzero-shot learningnode reconstructioncross-domain generalizationspectral normalizationclustering modulediscrepancy scoring

0 comments

The pith

AlignGAD performs zero-shot graph anomaly detection by aligning features across domains and scoring node reconstruction discrepancies.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper presents AlignGAD as a framework for identifying abnormal nodes in graphs from unseen domains without any training data from the target domain. Current approaches are limited because they depend on dataset-specific patterns that do not transfer. AlignGAD addresses this through unification of node features, construction of cluster views, and discrepancy scoring based on reconstruction. This matters because it opens the possibility of applying anomaly detection to new graph datasets immediately upon arrival. Experiments across real-world datasets support its effectiveness in the zero-shot setting.

Core claim

AlignGAD consists of a Global Unification Module to align heterogeneous node features and normalize graph signals in the spectral domain, a Clustering Module to construct cluster-aware graph views for capturing group-level patterns, and a Node Discrepancy Scoring Module to measure reconstruction discrepancy and aggregate anomaly evidence, allowing effective anomaly detection on unseen target graphs.

What carries the argument

The Global Unification Module that aligns heterogeneous node features and normalizes graph signals in the spectral domain to enable cross-domain transfer.

If this is right

Zero-shot GAD becomes feasible on new graphs without domain-specific retraining.
Node reconstruction discrepancies can serve as reliable anomaly indicators across different graph structures.
Cluster-aware views help capture abnormal patterns at the group level in heterogeneous data.
Generalization across domains reduces the need for labeled data in each new graph application.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

Similar unification techniques could apply to other cross-domain graph tasks such as classification or link prediction.
Success would imply that spectral domain normalization captures transferable anomaly signals better than raw features.
The framework might be extended by incorporating additional view constructions beyond clustering.

Load-bearing premise

The Global Unification Module can align heterogeneous node features and normalize graph signals in the spectral domain without losing critical information needed to detect anomalies in unseen target graphs.

What would settle it

A counterexample would be a target graph where AlignGAD's anomaly scores do not correlate with actual anomalies despite successful feature alignment, or where it performs no better than a random baseline.

Figures

Figures reproduced from arXiv: 2606.12673 by Dat Cao, Hien Chu, Khue Hoang, Phan Nguyen.

**Figure 1.** Figure 1: Difference between conventional anomaly detection and graph anomaly detection. Conventional anomaly detection focuses on isolated objects in feature space, while graph anomaly detection considers both node attributes and graph relations to identify abnormal nodes and abnormal connections. making direct feature transfer unreliable. Second, anomalies are not always expressed at the individual node level. I… view at source ↗

**Figure 2.** Figure 2: Comparison of GAD settings. Conventional supervised and unsupervised GAD methods train and infer within the same domain, while generalized GAD trains on source graphs and directly detects anomalies in unseen target graphs. Graph Anomaly Detection. Graph Anomaly Detection (GAD) aims to identify nodes or local graph structures that exhibit abnormal behaviors compared to the majority of the graph [23, 16, 8]… view at source ↗

**Figure 3.** Figure 3: The overall framework of AlignGAD. The framework consists of three modules: Global Unification Module, Clustering Module, and Node Discrepancy Scoring Module. The anomaly score is obtained by aggregating the outputs from different graph views. Feature Dimension Alignment. Graphs from different domains often contain node features with inconsistent dimensions, making direct cross-domain learning difficult. … view at source ↗

read the original abstract

Cross-domain graph anomaly detection (GAD) aims to identify abnormal nodes in unseen target graphs, showing strong potential in real-world applications with heterogeneous graph data. However, existing methods often depend on dataset-specific feature semantics and structural patterns, which limits their ability to generalize across different domains. To address this challenge, we propose AlignGAD, a zero-shot generalized graph anomaly detection framework. Our framework is built upon three key components: a Global Unification Module that aligns heterogeneous node features and normalizes graph signals in the spectral domain; a Clustering Module that constructs cluster-aware graph views to capture group-level abnormal patterns; and a Node Discrepancy Scoring Module that measures reconstruction discrepancy and aggregates anomaly evidence from different graph views. Experiments on multiple real-world datasets demonstrate the effectiveness of AlignGAD under the zero-shot GAD setting.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Desk Editor's Note private letter to a colleague

The abstract sketches AlignGAD with three modules for zero-shot cross-domain GAD but supplies zero numbers or implementation details, so the effectiveness claim cannot be checked.

read the letter

The paper's core pitch is a framework called AlignGAD that tries to do graph anomaly detection on unseen target graphs without any labeled or even unlabeled data from the target domain. It breaks the task into a Global Unification Module that aligns node features and does spectral normalization, a Clustering Module that makes cluster-aware views to pick up group-level anomalies, and a Node Discrepancy Scoring Module that scores reconstruction error across those views.

What is actually new is the explicit zero-shot framing plus the combination of spectral alignment with multi-view clustering for anomaly aggregation. The problem statement is clear: most prior GAD work ties itself to one dataset's feature semantics and structure, so moving to a new domain breaks it. The three-module structure is a reasonable way to try to break that tie.

The paper does a fair job naming the practical need. Real-world graphs often come from different sources with mismatched features, so a method that claims to ignore the target domain entirely would be useful if it held up.

The soft spot is obvious and central: the abstract says experiments on multiple real-world datasets demonstrate effectiveness, yet it gives no numbers, no baselines, no tables, and no description of how the modules are implemented or trained. Without those, the claim that the unification step preserves the signals needed for anomaly detection stays untested. The assumption that spectral normalization plus clustering will transfer anomaly patterns across domains is stated but not evidenced here.

This is for graph-ML people already working on anomaly detection who want ideas for handling domain shift. A reader could pull the module names and try similar ideas, but only after seeing the actual code or results.

On current evidence the paper is not ready for serious refereeing. The full manuscript would need to show concrete metrics, ablations, and comparisons before it earns that time.

Referee Report

1 major / 0 minor

Summary. The manuscript proposes AlignGAD, a zero-shot generalized graph anomaly detection framework consisting of three modules: a Global Unification Module that aligns heterogeneous node features and normalizes graph signals in the spectral domain, a Clustering Module that constructs cluster-aware graph views to capture group-level abnormal patterns, and a Node Discrepancy Scoring Module that measures reconstruction discrepancy and aggregates anomaly evidence from different graph views. The central claim is that experiments on multiple real-world datasets demonstrate the effectiveness of AlignGAD under the zero-shot GAD setting.

Significance. Cross-domain zero-shot GAD addresses a practically relevant challenge in handling heterogeneous graph data without domain-specific retraining. If the framework's modules enable reliable generalization while preserving anomaly signals, the work could contribute to broader applicability of GAD methods. However, the provided text supplies no quantitative results, baselines, error bars, implementation details, or ablation studies, preventing assessment of whether the claims are supported or whether the approach advances the state of the art.

major comments (1)

Abstract: the claim that 'experiments on multiple real-world datasets demonstrate the effectiveness' is unsupported because the text supplies no quantitative results, baselines, error bars, dataset descriptions, or performance metrics, making it impossible to evaluate whether the data supports the central claim.

Simulated Author's Rebuttal

1 responses · 0 unresolved

We thank the referee for the detailed review and the identification of the unsupported claim in the abstract. We address this point directly below.

read point-by-point responses

Referee: Abstract: the claim that 'experiments on multiple real-world datasets demonstrate the effectiveness' is unsupported because the text supplies no quantitative results, baselines, error bars, dataset descriptions, or performance metrics, making it impossible to evaluate whether the data supports the central claim.

Authors: We agree with the referee that the abstract claim is currently unsupported in the provided manuscript text. The current version lacks a dedicated experiments section containing quantitative results, baselines, error bars, dataset descriptions, and performance metrics. We will revise the manuscript to incorporate a full experimental evaluation on multiple real-world datasets under the zero-shot GAD setting, including all requested details, to substantiate the abstract claim. revision: yes

Circularity Check

0 steps flagged

No significant circularity detected

full rationale

The abstract and available text describe AlignGAD via three high-level modules (Global Unification, Clustering, Node Discrepancy Scoring) without any equations, derivations, fitted parameters, or self-citations. No load-bearing step reduces a claimed prediction or result to its own inputs by construction, and no uniqueness theorem or ansatz is invoked. The zero-shot GAD claim rests on experimental effectiveness statements that are not mathematically derived in the visible content, leaving the derivation chain empty of circular reductions.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Only the abstract is available; no free parameters, axioms, or invented entities can be identified from the provided text.

pith-pipeline@v0.9.1-grok · 5668 in / 1029 out tokens · 25927 ms · 2026-06-27T10:10:15.884203+00:00 · methodology

discussion (0)

Reference graph

Works this paper leans on

30 extracted references · 19 canonical work pages

[1]

In: Bengio, Y., LeCun, Y

Bruna, J., Zaremba, W., Szlam, A., LeCun, Y.: Spectral networks and locally connected networks on graphs. In: Bengio, Y., LeCun, Y. (eds.) 2nd International Conference on Learning Representations, ICLR 2014, Banff, AB, Canada, April 14- 16, 2014, Conference Track Proceedings (2014), http://arxiv.org/abs/1312.6203

Pith/arXiv arXiv 2014
[2]

In: Proceedings of the Thirty-Third International Joint Conference on Artificial Intelligence

Cai, J., Zhang, Y., Fan, J., Ng, S.K.: Lg-fgad: an effective federated graph anomaly detection framework. In: Proceedings of the Thirty-Third International Joint Conference on Artificial Intelligence. IJCAI ’24 (2024). https://doi.org/10.24963/ijcai.2024/416, https://doi.org/10.24963/ijcai.2024/416

work page doi:10.24963/ijcai.2024/416 2024
[3]

In: The Twelfth International Conference on Learning Representations (2024), https://openreview.net/forum?id=CanomFZssu

Chen, J., Zhu, G., Yuan, C., Huang, Y.: Boosting graph anomaly detection with adaptive message passing. In: The Twelfth International Conference on Learning Representations (2024), https://openreview.net/forum?id=CanomFZssu

2024
[4]

Ding, K., Li, J., Bhanushali, R., Liu, H.: Deep Anomaly Detection on At- tributed Networks, pp. 594–602. https://doi.org/10.1137/1.9781611975673.67, https://epubs.siam.org/doi/abs/10.1137/1.9781611975673.67

work page doi:10.1137/1.9781611975673.67
[5]

IEEE Transactions on Neural Networks and Learning Systems33(6), 2406–2415 (2022)

Ding, K., Shu, K., Shan, X., Li, J., Liu, H.: Cross-domain graph anomaly detection. IEEE Transactions on Neural Networks and Learning Systems33(6), 2406–2415 (2022). https://doi.org/10.1109/TNNLS.2021.3110982 AlignGAD: A Zero-shot Generalized GAD Framework 13

work page doi:10.1109/tnnls.2021.3110982 2022
[6]

Specaugment on large scale datasets

Fan, H., Zhang, F., Li, Z.: Anomalydae: Dual autoencoder for anomaly detec- tion on attributed networks. In: 2020 IEEE International Conference on Acoustics, Speech and Signal Processing, ICASSP 2020, Barcelona, Spain, May 4-8, 2020. pp. 5685–5689. IEEE (2020). https://doi.org/10.1109/ICASSP40776.2020.9053387, https://doi.org/10.1109/ICASSP40776.2020.9053387

work page doi:10.1109/icassp40776.2020.9053387 2020
[7]

In: Proceedings of the ACM Web Conference 2023

Gao, Y., Wang, X., He, X., Liu, Z., Feng, H., Zhang, Y.: Addressing heterophily in graph anomaly detection: A perspective of graph spectrum. In: Proceedings of the ACM Web Conference 2023. p. 1528–1538. WWW ’23, Association for Computing Machinery,NewYork,NY,USA(2023).https://doi.org/10.1145/3543507.3583268, https://doi.org/10.1145/3543507.3583268

work page doi:10.1145/3543507.3583268 2023
[8]

In: Proceed- ings of the Sixteenth ACM International Conference on Web Search and Data Mining

Gao, Y., Wang, X., He, X., Liu, Z., Feng, H., Zhang, Y.: Alleviat- ing structural distribution shift in graph anomaly detection. In: Proceed- ings of the Sixteenth ACM International Conference on Web Search and Data Mining. p. 357–365. WSDM ’23, Association for Computing Machin- ery, New York, NY, USA (2023). https://doi.org/10.1145/3539597.3570377, https...

work page doi:10.1145/3539597.3570377 2023
[9]

IEEE Trans

Guo, J., Tang, S., Li, J., Pan, K., Wu, L.: Rustgraph: Robust anomaly detection in dynamic graphs by jointly learning structural- temporal dependency. IEEE Trans. on Knowl. and Data Eng.36(7), 3472–3485 (Jul 2024). https://doi.org/10.1109/TKDE.2023.3328645, https://doi.org/10.1109/TKDE.2023.3328645

work page doi:10.1109/tkde.2023.3328645 2024
[10]

IEEE Transactions on Information Forensics and Security19, 8760–8772 (2024)

He, S., Li, G., Xie, K., Sharma, P.K.: Fusion graph structure learning-based multivariate time series anomaly detection with structured prior knowledge. IEEE Transactions on Information Forensics and Security19, 8760–8772 (2024). https://doi.org/10.1109/TIFS.2024.3459631

work page doi:10.1109/tifs.2024.3459631 2024
[11]

Ivanov, S., Prokhorenkova, L.: Boost then convolve: Gradient boosting meets graph neural networks (2021), https://arxiv.org/abs/2101.08543

arXiv 2021
[12]

Webb, Irwin King, and Shirui Pan

Jin, M., Koh, H.Y., Wen, Q., Zambon, D., Alippi, C., Webb, G.I., King, I., Pan, S.: A Survey on Graph Neural Networks for Time Se- ries: Forecasting, Classification, Imputation, and Anomaly Detection . IEEE Transactions on Pattern Analysis & Machine Intelligence46(12), 10466–10485 (Dec 2024). https://doi.org/10.1109/TPAMI.2024.3443141, https://doi.ieeecom...

work page doi:10.1109/tpami.2024.3443141 2024
[13]

In: 5th International Conference on Learning Representations, ICLR 2017, Toulon, France, April 24-26, 2017, Conference Track Proceedings

Kipf, T.N., Welling, M.: Semi-supervised classification with graph convolutional networks. In: 5th International Conference on Learning Representations, ICLR 2017, Toulon, France, April 24-26, 2017, Conference Track Proceedings. OpenRe- view.net (2017), https://openreview.net/forum?id=SJU4ayYgl

2017
[14]

In: Globerson, A., Mackey, L., Bel- grave, D., Fan, A., Paquet, U., Tomczak, J., Zhang, C

Liu, Y., Li, S., Zheng, Y., Chen, Q., Zhang, C., Pan, S.: Arc: A generalist graph anomaly detector with in-context learning. In: Globerson, A., Mackey, L., Bel- grave, D., Fan, A., Paquet, U., Tomczak, J., Zhang, C. (eds.) Advances in Neural Information Processing Systems. vol. 37, pp. 50772–50804. Curran Associates, Inc. (2024). https://doi.org/10.52202/...

work page doi:10.52202/079017-1606 2024
[15]

IEEE Transactions on Neural Networks and Learning Systems 33(6), 2378–2392 (2022)

Liu, Y., Li, Z., Pan, S., Gong, C., Zhou, C., Karypis, G.: Anomaly detection on attributed networks via contrastive self-supervised learn- ing. IEEE Transactions on Neural Networks and Learning Systems 33(6), 2378–2392 (2022). https://doi.org/10.1109/tnnls.2021.3068344, http://dx.doi.org/10.1109/TNNLS.2021.3068344

work page doi:10.1109/tnnls.2021.3068344 2022
[16]

In: Proceed- ings of the Fifteenth ACM International Conference on Web Search and 14 Phan Nguyen et al

Luo, X., Wu, J., Beheshti, A., Yang, J., Zhang, X., Wang, Y., Xue, S.: Comga: Community-aware attributed graph anomaly detection. In: Proceed- ings of the Fifteenth ACM International Conference on Web Search and 14 Phan Nguyen et al. Data Mining. p. 657–665. WSDM ’22, Association for Computing Machin- ery, New York, NY, USA (2022). https://doi.org/10.1145...

work page doi:10.1145/3488560.3498389 2022
[17]

In: Kwok, J

Niu, C., Qiao, H., Chen, C., Chen, L., Pang, G.: Zero-shot generalist graph anomaly detection with unified neighborhood prompts. In: Kwok, J. (ed.) Pro- ceedings of the Thirty-Fourth International Joint Conference on Artificial In- telligence, IJCAI-25. pp. 3226–3234. International Joint Conferences on Artifi- cial Intelligence Organization (8 2025). http...

work page doi:10.24963/ijcai.2025/359 2025
[18]

Proceedings of the AAAI Conference on Artificial Intelligence 38(21), 23610–23612 (Mar 2024)

Pang, S., Xiao, C., Tai, W., Cheng, Z., Zhou, F.: Graph anomaly de- tection with diffusion model-based graph enhancement (student ab- stract). Proceedings of the AAAI Conference on Artificial Intelligence 38(21), 23610–23612 (Mar 2024). https://doi.org/10.1609/aaai.v38i21.30494, https://ojs.aaai.org/index.php/AAAI/article/view/30494

work page doi:10.1609/aaai.v38i21.30494 2024
[19]

IEEE Trans

Pazho, A.D., Noghre, G.A., Purkayastha, A.A., Vempati, J., Martin, O., Tabkhi, H.: A survey of graph-based deep learning for anomaly detection in distributed systems. IEEE Trans. on Knowl. and Data Eng.36(1), 1–20 (Jan 2024). https://doi.org/10.1109/TKDE.2023.3282898, https://doi.org/10.1109/TKDE.2023.3282898

work page doi:10.1109/tkde.2023.3282898 2024
[20]

In: Proceedings of the 37th International Conference on Neural Information Processing Systems

Qiao,H.,Pang,G.:Truncatedaffinitymaximization:one-classhomophilymodeling for graph anomaly detection. In: Proceedings of the 37th International Conference on Neural Information Processing Systems. NIPS ’23, Curran Associates Inc., Red Hook, NY, USA (2023)

2023
[21]

In: Proceedings of the 38th International Conference on Neural Information Processing Systems

Qiao, H., Wen, Q., Li, X., Lim, E.P., Pang, G.: Generative semi-supervised graph anomaly detection. In: Proceedings of the 38th International Conference on Neural Information Processing Systems. NIPS ’24, Curran Associates Inc., Red Hook, NY, USA (2024)

2024
[22]

In: The Thirty-eighth Annual Conference on Neural Informa- tion Processing Systems (2024), https://openreview.net/forum?id=zqLAMwVLkt

Qiao, H., Wen, Q., Li, X., Lim, E.P., Pang, G.: Generative semi-supervised graph anomaly detection. In: The Thirty-eighth Annual Conference on Neural Informa- tion Processing Systems (2024), https://openreview.net/forum?id=zqLAMwVLkt

2024
[23]

In: International Conference on Machine Learning (2022)

Tang, J., Li, J., Gao, Z., Li, J.: Rethinking graph neural networks for anomaly detection. In: International Conference on Machine Learning (2022)

2022
[24]

In: 6th International Conference on Learning Representations, ICLR 2018, Vancouver,BC, Canada, April30 -May 3,2018, ConferenceTrack Pro- ceedings

Velickovic, P., Cucurull, G., Casanova, A., Romero, A., Liò, P., Bengio, Y.: Graph attention networks. In: 6th International Conference on Learning Representations, ICLR 2018, Vancouver,BC, Canada, April30 -May 3,2018, ConferenceTrack Pro- ceedings. OpenReview.net (2018), https://openreview.net/forum?id=rJXMpikCZ

2018
[25]

AAAI’23/IAAI’23/EAAI’23, AAAI Press (2023)

Wang, Q., Pang, G., Salehi, M., Buntine, W., Leckie, C.: Cross-domain graph anomaly detection via anomaly- aware contrastive alignment. AAAI’23/IAAI’23/EAAI’23, AAAI Press (2023). https://doi.org/10.1609/aaai.v37i4.25591, https://doi.org/10.1609/aaai.v37i4.25591

work page doi:10.1609/aaai.v37i4.25591 2023
[26]

In: Proceedings of the Thirty- Third International Joint Conference on Artificial Intelligence

Wu, J., Hu, R., Li, D., Huang, Z., Ren, L., Zang, Y.: Robust heterophilic graph learning against label noise for anomaly detection. In: Proceedings of the Thirty- Third International Joint Conference on Artificial Intelligence. IJCAI ’24 (2024). https://doi.org/10.24963/ijcai.2024/271, https://doi.org/10.24963/ijcai.2024/271

work page doi:10.24963/ijcai.2024/271 2024
[27]

Xiao, C., Pang, S., Xu, X., Li, X., Trajcevski, G., Zhou, F.: Counterfactual data augmentation with denoising diffusion for graph anomaly detection (2024), https://arxiv.org/abs/2407.02143

arXiv 2024
[28]

In: Proceedings of the Thirty-Fourth In- ternational Joint Conference on Artificial Intelligence

Zhang, X., Peng, H., He, Z., Xie, C., Jin, X., Jiang, H.: Gctam: global and contextual truncated affinity combined maximization model for unsu- AlignGAD: A Zero-shot Generalized GAD Framework 15 pervised graph anomaly detection. In: Proceedings of the Thirty-Fourth In- ternational Joint Conference on Artificial Intelligence. IJCAI ’25 (2025). https://doi....

work page doi:10.24963/ijcai.2025/405 2025
[29]

IEEE Trans

Zheng, B., Ming, L., Zeng, K., Zhou, M., Zhang, X., Ye, T., Yang, B., Zhou, X., Jensen, C.S.: Adversarial graph neural network for multivari- ate time series anomaly detection. IEEE Trans. on Knowl. and Data Eng. 36(12), 7612–7626 (Dec 2024). https://doi.org/10.1109/TKDE.2024.3419891, https://doi.org/10.1109/TKDE.2024.3419891

work page doi:10.1109/tkde.2024.3419891 2024
[30]

IEEE Transactions on Neural Networks and Learning Systems (TNNLS) (2023)

Zheng, Y., Koh, H.Y., Jin, M., Chi, L., Phan, K.T., Pan, S., Chen, Y.P.P., Xiang, W.: Correlation-aware Spatial-Temporal Graph Learning for Multivariate Time- series Anomaly Detection. IEEE Transactions on Neural Networks and Learning Systems (TNNLS) (2023)

2023

[1] [1]

In: Bengio, Y., LeCun, Y

Bruna, J., Zaremba, W., Szlam, A., LeCun, Y.: Spectral networks and locally connected networks on graphs. In: Bengio, Y., LeCun, Y. (eds.) 2nd International Conference on Learning Representations, ICLR 2014, Banff, AB, Canada, April 14- 16, 2014, Conference Track Proceedings (2014), http://arxiv.org/abs/1312.6203

Pith/arXiv arXiv 2014

[2] [2]

In: Proceedings of the Thirty-Third International Joint Conference on Artificial Intelligence

Cai, J., Zhang, Y., Fan, J., Ng, S.K.: Lg-fgad: an effective federated graph anomaly detection framework. In: Proceedings of the Thirty-Third International Joint Conference on Artificial Intelligence. IJCAI ’24 (2024). https://doi.org/10.24963/ijcai.2024/416, https://doi.org/10.24963/ijcai.2024/416

work page doi:10.24963/ijcai.2024/416 2024

[3] [3]

In: The Twelfth International Conference on Learning Representations (2024), https://openreview.net/forum?id=CanomFZssu

Chen, J., Zhu, G., Yuan, C., Huang, Y.: Boosting graph anomaly detection with adaptive message passing. In: The Twelfth International Conference on Learning Representations (2024), https://openreview.net/forum?id=CanomFZssu

2024

[4] [4]

Ding, K., Li, J., Bhanushali, R., Liu, H.: Deep Anomaly Detection on At- tributed Networks, pp. 594–602. https://doi.org/10.1137/1.9781611975673.67, https://epubs.siam.org/doi/abs/10.1137/1.9781611975673.67

work page doi:10.1137/1.9781611975673.67

[5] [5]

IEEE Transactions on Neural Networks and Learning Systems33(6), 2406–2415 (2022)

Ding, K., Shu, K., Shan, X., Li, J., Liu, H.: Cross-domain graph anomaly detection. IEEE Transactions on Neural Networks and Learning Systems33(6), 2406–2415 (2022). https://doi.org/10.1109/TNNLS.2021.3110982 AlignGAD: A Zero-shot Generalized GAD Framework 13

work page doi:10.1109/tnnls.2021.3110982 2022

[6] [6]

Specaugment on large scale datasets

Fan, H., Zhang, F., Li, Z.: Anomalydae: Dual autoencoder for anomaly detec- tion on attributed networks. In: 2020 IEEE International Conference on Acoustics, Speech and Signal Processing, ICASSP 2020, Barcelona, Spain, May 4-8, 2020. pp. 5685–5689. IEEE (2020). https://doi.org/10.1109/ICASSP40776.2020.9053387, https://doi.org/10.1109/ICASSP40776.2020.9053387

work page doi:10.1109/icassp40776.2020.9053387 2020

[7] [7]

In: Proceedings of the ACM Web Conference 2023

Gao, Y., Wang, X., He, X., Liu, Z., Feng, H., Zhang, Y.: Addressing heterophily in graph anomaly detection: A perspective of graph spectrum. In: Proceedings of the ACM Web Conference 2023. p. 1528–1538. WWW ’23, Association for Computing Machinery,NewYork,NY,USA(2023).https://doi.org/10.1145/3543507.3583268, https://doi.org/10.1145/3543507.3583268

work page doi:10.1145/3543507.3583268 2023

[8] [8]

In: Proceed- ings of the Sixteenth ACM International Conference on Web Search and Data Mining

Gao, Y., Wang, X., He, X., Liu, Z., Feng, H., Zhang, Y.: Alleviat- ing structural distribution shift in graph anomaly detection. In: Proceed- ings of the Sixteenth ACM International Conference on Web Search and Data Mining. p. 357–365. WSDM ’23, Association for Computing Machin- ery, New York, NY, USA (2023). https://doi.org/10.1145/3539597.3570377, https...

work page doi:10.1145/3539597.3570377 2023

[9] [9]

IEEE Trans

Guo, J., Tang, S., Li, J., Pan, K., Wu, L.: Rustgraph: Robust anomaly detection in dynamic graphs by jointly learning structural- temporal dependency. IEEE Trans. on Knowl. and Data Eng.36(7), 3472–3485 (Jul 2024). https://doi.org/10.1109/TKDE.2023.3328645, https://doi.org/10.1109/TKDE.2023.3328645

work page doi:10.1109/tkde.2023.3328645 2024

[10] [10]

IEEE Transactions on Information Forensics and Security19, 8760–8772 (2024)

He, S., Li, G., Xie, K., Sharma, P.K.: Fusion graph structure learning-based multivariate time series anomaly detection with structured prior knowledge. IEEE Transactions on Information Forensics and Security19, 8760–8772 (2024). https://doi.org/10.1109/TIFS.2024.3459631

work page doi:10.1109/tifs.2024.3459631 2024

[11] [11]

Ivanov, S., Prokhorenkova, L.: Boost then convolve: Gradient boosting meets graph neural networks (2021), https://arxiv.org/abs/2101.08543

arXiv 2021

[12] [12]

Webb, Irwin King, and Shirui Pan

Jin, M., Koh, H.Y., Wen, Q., Zambon, D., Alippi, C., Webb, G.I., King, I., Pan, S.: A Survey on Graph Neural Networks for Time Se- ries: Forecasting, Classification, Imputation, and Anomaly Detection . IEEE Transactions on Pattern Analysis & Machine Intelligence46(12), 10466–10485 (Dec 2024). https://doi.org/10.1109/TPAMI.2024.3443141, https://doi.ieeecom...

work page doi:10.1109/tpami.2024.3443141 2024

[13] [13]

In: 5th International Conference on Learning Representations, ICLR 2017, Toulon, France, April 24-26, 2017, Conference Track Proceedings

Kipf, T.N., Welling, M.: Semi-supervised classification with graph convolutional networks. In: 5th International Conference on Learning Representations, ICLR 2017, Toulon, France, April 24-26, 2017, Conference Track Proceedings. OpenRe- view.net (2017), https://openreview.net/forum?id=SJU4ayYgl

2017

[14] [14]

In: Globerson, A., Mackey, L., Bel- grave, D., Fan, A., Paquet, U., Tomczak, J., Zhang, C

Liu, Y., Li, S., Zheng, Y., Chen, Q., Zhang, C., Pan, S.: Arc: A generalist graph anomaly detector with in-context learning. In: Globerson, A., Mackey, L., Bel- grave, D., Fan, A., Paquet, U., Tomczak, J., Zhang, C. (eds.) Advances in Neural Information Processing Systems. vol. 37, pp. 50772–50804. Curran Associates, Inc. (2024). https://doi.org/10.52202/...

work page doi:10.52202/079017-1606 2024

[15] [15]

IEEE Transactions on Neural Networks and Learning Systems 33(6), 2378–2392 (2022)

Liu, Y., Li, Z., Pan, S., Gong, C., Zhou, C., Karypis, G.: Anomaly detection on attributed networks via contrastive self-supervised learn- ing. IEEE Transactions on Neural Networks and Learning Systems 33(6), 2378–2392 (2022). https://doi.org/10.1109/tnnls.2021.3068344, http://dx.doi.org/10.1109/TNNLS.2021.3068344

work page doi:10.1109/tnnls.2021.3068344 2022

[16] [16]

In: Proceed- ings of the Fifteenth ACM International Conference on Web Search and 14 Phan Nguyen et al

Luo, X., Wu, J., Beheshti, A., Yang, J., Zhang, X., Wang, Y., Xue, S.: Comga: Community-aware attributed graph anomaly detection. In: Proceed- ings of the Fifteenth ACM International Conference on Web Search and 14 Phan Nguyen et al. Data Mining. p. 657–665. WSDM ’22, Association for Computing Machin- ery, New York, NY, USA (2022). https://doi.org/10.1145...

work page doi:10.1145/3488560.3498389 2022

[17] [17]

In: Kwok, J

Niu, C., Qiao, H., Chen, C., Chen, L., Pang, G.: Zero-shot generalist graph anomaly detection with unified neighborhood prompts. In: Kwok, J. (ed.) Pro- ceedings of the Thirty-Fourth International Joint Conference on Artificial In- telligence, IJCAI-25. pp. 3226–3234. International Joint Conferences on Artifi- cial Intelligence Organization (8 2025). http...

work page doi:10.24963/ijcai.2025/359 2025

[18] [18]

Proceedings of the AAAI Conference on Artificial Intelligence 38(21), 23610–23612 (Mar 2024)

Pang, S., Xiao, C., Tai, W., Cheng, Z., Zhou, F.: Graph anomaly de- tection with diffusion model-based graph enhancement (student ab- stract). Proceedings of the AAAI Conference on Artificial Intelligence 38(21), 23610–23612 (Mar 2024). https://doi.org/10.1609/aaai.v38i21.30494, https://ojs.aaai.org/index.php/AAAI/article/view/30494

work page doi:10.1609/aaai.v38i21.30494 2024

[19] [19]

IEEE Trans

Pazho, A.D., Noghre, G.A., Purkayastha, A.A., Vempati, J., Martin, O., Tabkhi, H.: A survey of graph-based deep learning for anomaly detection in distributed systems. IEEE Trans. on Knowl. and Data Eng.36(1), 1–20 (Jan 2024). https://doi.org/10.1109/TKDE.2023.3282898, https://doi.org/10.1109/TKDE.2023.3282898

work page doi:10.1109/tkde.2023.3282898 2024

[20] [20]

In: Proceedings of the 37th International Conference on Neural Information Processing Systems

Qiao,H.,Pang,G.:Truncatedaffinitymaximization:one-classhomophilymodeling for graph anomaly detection. In: Proceedings of the 37th International Conference on Neural Information Processing Systems. NIPS ’23, Curran Associates Inc., Red Hook, NY, USA (2023)

2023

[21] [21]

In: Proceedings of the 38th International Conference on Neural Information Processing Systems

Qiao, H., Wen, Q., Li, X., Lim, E.P., Pang, G.: Generative semi-supervised graph anomaly detection. In: Proceedings of the 38th International Conference on Neural Information Processing Systems. NIPS ’24, Curran Associates Inc., Red Hook, NY, USA (2024)

2024

[22] [22]

In: The Thirty-eighth Annual Conference on Neural Informa- tion Processing Systems (2024), https://openreview.net/forum?id=zqLAMwVLkt

Qiao, H., Wen, Q., Li, X., Lim, E.P., Pang, G.: Generative semi-supervised graph anomaly detection. In: The Thirty-eighth Annual Conference on Neural Informa- tion Processing Systems (2024), https://openreview.net/forum?id=zqLAMwVLkt

2024

[23] [23]

In: International Conference on Machine Learning (2022)

Tang, J., Li, J., Gao, Z., Li, J.: Rethinking graph neural networks for anomaly detection. In: International Conference on Machine Learning (2022)

2022

[24] [24]

In: 6th International Conference on Learning Representations, ICLR 2018, Vancouver,BC, Canada, April30 -May 3,2018, ConferenceTrack Pro- ceedings

Velickovic, P., Cucurull, G., Casanova, A., Romero, A., Liò, P., Bengio, Y.: Graph attention networks. In: 6th International Conference on Learning Representations, ICLR 2018, Vancouver,BC, Canada, April30 -May 3,2018, ConferenceTrack Pro- ceedings. OpenReview.net (2018), https://openreview.net/forum?id=rJXMpikCZ

2018

[25] [25]

AAAI’23/IAAI’23/EAAI’23, AAAI Press (2023)

Wang, Q., Pang, G., Salehi, M., Buntine, W., Leckie, C.: Cross-domain graph anomaly detection via anomaly- aware contrastive alignment. AAAI’23/IAAI’23/EAAI’23, AAAI Press (2023). https://doi.org/10.1609/aaai.v37i4.25591, https://doi.org/10.1609/aaai.v37i4.25591

work page doi:10.1609/aaai.v37i4.25591 2023

[26] [26]

In: Proceedings of the Thirty- Third International Joint Conference on Artificial Intelligence

Wu, J., Hu, R., Li, D., Huang, Z., Ren, L., Zang, Y.: Robust heterophilic graph learning against label noise for anomaly detection. In: Proceedings of the Thirty- Third International Joint Conference on Artificial Intelligence. IJCAI ’24 (2024). https://doi.org/10.24963/ijcai.2024/271, https://doi.org/10.24963/ijcai.2024/271

work page doi:10.24963/ijcai.2024/271 2024

[27] [27]

Xiao, C., Pang, S., Xu, X., Li, X., Trajcevski, G., Zhou, F.: Counterfactual data augmentation with denoising diffusion for graph anomaly detection (2024), https://arxiv.org/abs/2407.02143

arXiv 2024

[28] [28]

In: Proceedings of the Thirty-Fourth In- ternational Joint Conference on Artificial Intelligence

Zhang, X., Peng, H., He, Z., Xie, C., Jin, X., Jiang, H.: Gctam: global and contextual truncated affinity combined maximization model for unsu- AlignGAD: A Zero-shot Generalized GAD Framework 15 pervised graph anomaly detection. In: Proceedings of the Thirty-Fourth In- ternational Joint Conference on Artificial Intelligence. IJCAI ’25 (2025). https://doi....

work page doi:10.24963/ijcai.2025/405 2025

[29] [29]

IEEE Trans

Zheng, B., Ming, L., Zeng, K., Zhou, M., Zhang, X., Ye, T., Yang, B., Zhou, X., Jensen, C.S.: Adversarial graph neural network for multivari- ate time series anomaly detection. IEEE Trans. on Knowl. and Data Eng. 36(12), 7612–7626 (Dec 2024). https://doi.org/10.1109/TKDE.2024.3419891, https://doi.org/10.1109/TKDE.2024.3419891

work page doi:10.1109/tkde.2024.3419891 2024

[30] [30]

IEEE Transactions on Neural Networks and Learning Systems (TNNLS) (2023)

Zheng, Y., Koh, H.Y., Jin, M., Chi, L., Phan, K.T., Pan, S., Chen, Y.P.P., Xiang, W.: Correlation-aware Spatial-Temporal Graph Learning for Multivariate Time- series Anomaly Detection. IEEE Transactions on Neural Networks and Learning Systems (TNNLS) (2023)

2023