Pith. sign in

REVIEW 2 major objections 6 minor 118 references

SALT-GNN: Handling Dense Neighborhoods in Anti-Money Laundering Graphs via Statistics-Aware Attention

T0 review · 2 major / 6 minor · reviewed 2026-07-14 · grok-4.5

Pith's one-line read Dense recipient accounts hide laundering signals from standard graph neural nets; fusing degree-aware statistics with attention at every layer recovers them with a compact model.

desk verdict Solid AML-GNN paper: recipient-degree diagnostics plus layer-wise stats–attention fusion that actually holds up on the evidence they have. read the letter →

arxiv 2607.10131 v1 pith:BS27O6HH submitted 2026-07-11 cs.LG cs.AI

classification cs.LGcs.AI
keywords anti-moneylaunderinggraphneuralnetworksattentiondenseneighborhoodsrecipient-degreestratificationstatistics-awareaggregationlayer-wisefusionfinancialtransactiongraphs
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

Anti-money-laundering systems that use graph neural networks usually report one overall F1 score, which hides a practical failure: high-activity recipient accounts are exactly where suspicious signals are hardest to isolate and most expensive to investigate. The paper shows, with a post-hoc stratification of predictions by recipient in-degree, that performance of standard GNNs systematically falls in those dense contexts on three public benchmarks. It traces the fall to three aggregation limits that transaction graphs amplify—multiset non-discriminability, cardinality blindness, and attenuation of weak multi-hop evidence under normalized attention—and answers them with SALT-GNN: at each message-passing layer a degree-aware multi-statistic channel is fused with an attention channel so that later attention steps already see distributional and size cues. The same layer-wise fusion improves dense-bin F1 with far fewer parameters than task-specific graph transformers, and the gain appears for both Transformer-style and GAT-style attention, pointing to fusion placement rather than the choice of attention operator.

What carries the argument

SALT-GNN layer-wise fusion: at each message-passing step a residual combination of the previous node state, a degree-scaled multi-statistic summary (mean, std, min, max), and a learned attention aggregation is formed so that subsequent attention scores operate on statistics-enriched representations.

What would settle it

On a proprietary multi-institution transaction graph with real labels, retrain the same SALT variants and the strongest baselines under identical protocols and check whether the dense recipient-degree bins still show the reported F1 lift of several points over late-fusion and attention-only baselines.

Watch

Extended reading notes

Core claim

Recipient-degree stratified evaluation reveals consistent degradation of standard AML GNNs in dense recipient contexts; SALT-GNN, which fuses degree-aware statistical aggregation with attention at every message-passing layer, improves dense-context F1 by 3–6 points on HI-Small and HI-Medium and by 16–20 points in the highest-degree bin of AMLSim-32k-5%, using up to 77% fewer parameters than the strongest task-specific graph-transformer baselines, with the benefit coming from where the two evidence streams interact rather than from a particular attention operator.

Load-bearing premise

The public synthetic AML graphs and their planted fraud patterns are representative enough of real multi-institution transaction data that both the dense-bin degradation and the layer-wise fusion benefit will transfer to production systems.

Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

2 major / 6 minor

Summary. The paper argues that aggregate F1 hides systematic GNN degradation in dense recipient contexts on AML transaction graphs, and that this failure is driven by multiset non-discriminability, cardinality blindness, and attention-induced multi-hop attenuation. It introduces a post-hoc recipient-degree stratified evaluation protocol and SALT-GNN, which fuses degree-aware multi-statistic aggregation with attention at every message-passing layer so that distributional and cardinality cues shape subsequent attention states. Empirically, on HI-Small, HI-Medium, and AMLSim-32k-5%, SALT improves dense-bin F1 (about 3–6 points on IBM tasks; 16–20 in AMLSim’s extreme bin) with a compact parameter budget, for both Transformer- and GAT-style attention. Ablations, Hit@k diagnostics, a within-model fusion-placement knockout, and controlled synthetic C1–C3 tests are used to attribute gains to layer-wise fusion rather than branch presence or a particular attention operator.

Significance. If the results hold, the contribution is twofold and practically useful: (i) a simple, reusable evaluation unit—recipient in-degree—that makes dense-hub reliability a first-class AML-GNN reporting requirement, and (ii) a lightweight hybrid design principle (layer-wise statistics–attention fusion) that improves exactly those regimes without a heavy task-specific graph transformer. Strengths include multi-dataset coverage (edge and node tasks), degree-stratified F1 and PR-AUC, attention diagnostics (Hit@k, ENNs), characteristic-level ablations, a causal fusion-placement knockout (App. M), controlled synthetic isolation of C1–C3, Holm-corrected tests, and a promised code release for the stratified protocol and ablations. The main ceiling on impact is external validity: all evidence is on public synthetic AML graphs, which the authors acknowledge.

major comments (2)
  1. [§5, Tables 2–3; Appendix J] §5, Tables 2–3 and Appendix J: the headline dense-context gains on IBM (e.g., HI-Small deg[50–99], HI-Medium deg[100+]) rest on sparse positive support (on the order of ~124 / ~72 fraud cases in those test bins) together with n=3–5 seeds. After Holm correction several dense-bin contrasts are non-significant or only marginal, while effect sizes remain favorable. The central dense-bin claim is still directionally supported (especially by AMLSim’s paired tests and App. M), but the manuscript should surface per-bin positive counts next to the main stratified tables and state more carefully that the 3–6 point IBM dense gains are under-powered estimates rather than precisely established deltas.
  2. [§6; §4 Datasets] §6 and §4: external validity is correctly listed as a limitation, but it is load-bearing for the operational framing (investigation cost at high-activity recipients). The paper should expand, even briefly, which synthetic properties (agent-based/planted patterns, illicit-rate structure, single-generator topology, absence of multi-institution and product heterogeneity) could make either dense-bin degradation or the layer-wise fusion benefit fail to transfer, and what a minimal proprietary or multi-institution check would need to report under the same recipient-degree protocol. This does not require new private data in revision, but the transfer claim should be scoped more explicitly than the current short paragraph.
minor comments (6)
  1. [Abstract; §3] Abstract and §1: the three characteristics are clear, but Characteristic 3 is sometimes phrased as if SALT removes softmax attenuation; §3 correctly says it does not, and mitigates via a parallel statistical path. Align the abstract wording with that more precise claim.
  2. [Table 5; Appendix D] Table 5 / Appendix D: Hit@20 is well motivated by ENNs≈20 in deg[50–99], but the main text should state in one sentence why k is fixed rather than degree-adaptive, to avoid the impression that k was chosen post hoc for SALT.
  3. [§4; Appendix G, I] §4 Baselines / Appendix I: the hyperparameter provenance table is helpful. In the main experimental setup, add one sentence that SALT and PNAGMDA inherit backbone hyperparameters and that App. G reports dense-bin retuning of TransConv/FraudGT, so readers do not assume an unfair tuning advantage without checking the appendix.
  4. [§3; Figure 2; Appendix H] Figure 2 and Eqs. (9)–(11): the residual averaging h=(h_prev + h_fusion)/2 is easy to miss relative to the concat–MLP fusion. Cross-reference Appendix H in the main architecture section so residual design is not confused with the fusion-placement claim.
  5. [Throughout] Presentation: several places in the compiled text show concatenated words (e.g., “graphneuralnetworks”, “denserecipientcontexts”). Clean spacing and line-break artifacts throughout before camera-ready.
  6. [§2; §7] Related Work §7: dense-subgraph fraud methods are correctly called complementary; a one-sentence contrast with cardinality-preserving attention work (already cited Zhang & Xie 2021) in the main diagnosis section would help readers place Characteristic 2.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: empirical architecture paper whose dense-bin gains are measured on held-out splits, not derived by construction from fitted inputs or self-citation.

full rationale

SALT-GNN is an empirical GNN architecture paper. The load-bearing claims (recipient-degree degradation; layer-wise fusion of degree-aware multi-statistic aggregation with attention improving dense-context F1; fusion placement rather than attention operator as the driver) are established by post-hoc stratification of test predictions, external baselines (GIN, PNA, GAT, TransConv, FraudGT, PNAGMDA, GAMLNet), characteristic-level ablations, Hit@k diagnostics, and a within-model fusion-placement knockout on planted synthetic motifs (App. M). Degree scalers use the training-graph mean of log(d+1) exactly as in PNA (Corso et al. 2020)—a fixed normalization constant, not a parameter fitted to force dense-bin F1. Model selection is by overall validation F1; stratified metrics are reporting only. Citations for multiset non-discriminability, cardinality blindness, attention, and baselines are external (Corso, Zhang & Xie, Velickovic, Shi, Altman, Lin, Chen et al.); there is no self-citation uniqueness theorem or ansatz chain that forces the result. Performance numbers are not algebraically equivalent to any input definition. External-validity limits on synthetic AML graphs are stated as limitations, not circular reductions. Score 0 is therefore the correct finding.

Assumptions & free parameters 4 free parameters · 5 assumptions · 2 invented entities

Load-bearing content is mostly standard GNN math plus domain assumptions about synthetic AML graphs and operational cost of dense recipients. Free parameters are ordinary training hyperparameters and PNA degree-scaler normalization constants, not physics-style fitted constants that define the claim. Invented entities are methodological (evaluation protocol and architecture), not new physical objects.

free parameters (4)
  • PNA degree-scaler reference δ = mean log(d+1) on training graph
    Normalizes amplification/attenuation scalers; fixed from training graph statistics, inherited from PNA, and can shift absolute scale of the statistical channel across datasets.
  • Model hyperparameters (lr, hidden size, class weight, dropout, depth, heads)
    Selected by validation F1 or reused from prior AML configs; SALT inherits backbone settings. Performance claims depend on these choices though authors argue fusion, not extra tuning, drives gains.
  • Recipient in-degree bin edges (log-spaced ranges)
    Post-hoc stratification bins are design choices that define where ‘dense’ is reported; they do not train the model but structure the central diagnostic claim.
  • Hit@k threshold (k=20 on HI-Small deg[50–99])
    Chosen to match TransConv effective neighborhood size; affects attention-preservation narrative though not the F1 training objective.
assumptions (5)
  • standard math Common single-statistic neighborhood aggregators cannot uniquely characterize multisets (multiset non-discriminability).
    Invoked from Corso et al. 2020 / PNA literature in §2 Characteristic 1 as a known GNN limitation amplified by AML.
  • standard math Normalized or order-statistic aggregators and softmax attention do not by themselves expose neighborhood cardinality (cardinality blindness).
    Cited via Zhang & Xie 2021 and attention normalization in §2 Characteristic 2.
  • domain assumption Softmax attention influence compounds multiplicatively along multi-hop paths and can attenuate weak coordinated edges at dense hubs.
    Characteristic 3 in §2; follows from normalized attention math plus AML layering narrative; mechanism supported by Hit@k and synthetic routing tests but not a theorem unique to this paper.
  • domain assumption Synthetic IBM HI and AMLSim graphs with temporal/random splits are valid proxies for studying supervised AML GNN behavior under recipient density.
    Entire experimental program in §4–5 rests on these public benchmarks; authors note privacy-driven reliance on synthetic data.
  • ad hoc to paper Recipient in-degree is the right operational context unit for both edge (receiver) and node classification.
    Core evaluation design choice in §1 and §4; motivated operationally but not independently validated on production case-management costs.
invented entities (2)
  • Recipient-degree stratified evaluation protocol independent evidence
    purpose: Post-hoc binning of test predictions by recipient in-degree to expose dense-context degradation hidden by aggregate F1.
    Methodological construct introduced as Contribution 1; falsifiable via reimplementation on the same public graphs.
  • SALT-GNN (layer-wise statistics–attention fusion architecture)
    purpose: Fuse degree-aware multi-aggregators with attention each layer so subsequent attention operates on statistics-enriched states.
    Primary proposed model (SALT-Trans / SALT-GAT); evidence is empirical performance and ablations, not external physical measurement.

how reviews work

0 comments
Cite this review

Pith. "Pith review of SALT-GNN: Handling Dense Neighborhoods in Anti-Money Laundering Graphs via Statistics-Aware Attention." pith.science (2026). https://pith.science/paper/BS27O6HH

@misc{pith2026260710131,
  author       = {Pith},
  title        = {Pith review of: SALT-GNN: Handling Dense Neighborhoods in Anti-Money Laundering Graphs via Statistics-Aware Attention},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/BS27O6HH}},
  note         = {Machine review of arXiv:2607.10131}
}
read the original abstract

Money laundering threatens financial stability and exposes institutions to penalties, motivating automated detection. Because laundering schemes often emerge through relational patterns, graph neural networks (GNNs) are increasingly used for anti-money laundering (AML). Yet AML GNNs are typically evaluated with aggregate metrics such as overall F1 score, which hide an operational issue: high-activity recipient accounts concentrate many incoming transactions, making suspicious signals harder to isolate and costlier to investigate. We introduce a recipient-degree stratified evaluation that reports standard AML metrics across recipient-context density. Across three datasets (HI-Small, HI-Medium, and AMLSim-32k-5%), it reveals consistent degradation in dense recipient contexts, which we trace to three GNN characteristics: two known limitations that AML amplifies, i.e., (1) multiset non-discriminability and (2) cardinality blindness, and (3) an attention-specific effect: in dense neighborhoods, normalized attention attenuates weak but pattern-relevant multi-hop signals. Guided by this diagnosis, we propose SALT-GNN, a lightweight statistics-aware architecture that fuses degree-aware statistical aggregation with attention at each message-passing layer, so distributional and cardinality information shapes the node states used by subsequent attention steps. Ablations support fusion placement as a key factor in dense-context performance. On HI-Small and HI-Medium, SALT-GNN uses up to 77% fewer parameters than task-specific graph-transformer baselines while improving dense-context F1 score by 3-6 points; on AMLSim-32k-5%, it improves highest-degree F1 score by 16-20 points. The gains hold for both Transformer- and GAT-style attention, indicating that the benefit comes from where statistical and attentional evidence is fused rather than from a specific attention operator.

Figures

Figures reproduced from arXiv: 2607.10131 by the authors.

Figure 1
Figure 1. Recipient-context example. A suspicious incoming [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. SALT-GNN architecture. Two parallel branches [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Canonical money laundering patterns observed in [PITH_FULL_IMAGE:figures/full_fig_p009_3.png] view at source ↗
Figures from the paper (3 more)
Figure 4
Figure 4. Figure 4: Degree-stratified evaluation: from an overall metric [PITH_FULL_IMAGE:figures/full_fig_p009_4.png]
Figure 5
Figure 5. Figure 5: Average training time per epoch for HI-Small on a [PITH_FULL_IMAGE:figures/full_fig_p015_5.png]
Figure 6
Figure 6. Figure 6: A constructed two-hop fraud motif. A receiver [PITH_FULL_IMAGE:figures/full_fig_p018_6.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

118 extracted references · 4 canonical work pages

  1. [1]

    The Accounting Review , volume=

    The cost of fraud prediction errors , author=. The Accounting Review , volume=. 2022 , publisher=

  2. [2]

    Toyotaro Suzumura and Hiroki Kanezashi , title =

  3. [3]

    IEEE Transactions on Knowledge and Data Engineering , volume=

    A contrast metric for fraud detection in rich graphs , author=. IEEE Transactions on Knowledge and Data Engineering , volume=. 2018 , publisher=

  4. [4]

    Proceedings of the 25th ACM SIGKDD international conference on knowledge discovery & data mining , pages=

    Net: Degree-specific graph neural networks for node and graph classification , author=. Proceedings of the 25th ACM SIGKDD international conference on knowledge discovery & data mining , pages=

  5. [5]

    Advances in neural information processing systems , volume=

    Nodeformer: A scalable graph structure learning transformer for node classification , author=. Advances in neural information processing systems , volume=

  6. [6]

    Advances in Neural Information Processing Systems , volume=

    Relieving the Over-Aggregating Effect in Graph Transformers , author=. Advances in Neural Information Processing Systems , volume=

  7. [7]

    The Tenth International Conference on Learning Representations,

    Shaked Brody and Uri Alon and Eran Yahav , title =. The Tenth International Conference on Learning Representations,. 2022 , url =

  8. [8]

    ACM Transactions on Knowledge Discovery from Data (TKDD) , volume=

    Graph-based fraud detection in the face of camouflage , author=. ACM Transactions on Knowledge Discovery from Data (TKDD) , volume=. 2017 , publisher=

Show all 118 references
  1. [9]

    2024 11th IEEE Swiss Conference on Data Science (SDS) , pages=

    GAMLNet: a graph based framework for the detection of money laundering , author=. 2024 11th IEEE Swiss Conference on Data Science (SDS) , pages=. 2024 , organization=

  2. [10]

    Proceedings of the 31st ACM SIGKDD Conference on Knowledge Discovery and Data Mining V

    Feature Reconstruction for Anomaly Detection on Directed Multigraphs: A Preprocessing Framework for GNNs , author=. Proceedings of the 31st ACM SIGKDD Conference on Knowledge Discovery and Data Mining V. 2 , pages=

  3. [11]

    Proceedings of the 31st ACM SIGKDD Conference on Knowledge Discovery and Data Mining V

    MAD: Multicriteria Anomaly Detection of Suspicious Financial Accounts from Billions of Cash Transactions , author=. Proceedings of the 31st ACM SIGKDD Conference on Knowledge Discovery and Data Mining V. 2 , pages=

  4. [12]

    Proceedings of the 28th ACM SIGKDD Conference on Knowledge Discovery and Data Mining , pages=

    Accelerated GNN training with DGL and RAPIDS cuGraph in a Fraud Detection Workflow , author=. Proceedings of the 28th ACM SIGKDD Conference on Knowledge Discovery and Data Mining , pages=

  5. [13]

    Proceedings of the 31st ACM SIGKDD Conference on Knowledge Discovery and Data Mining V

    Flag: Fraud detection with llm-enhanced graph neural network , author=. Proceedings of the 31st ACM SIGKDD Conference on Knowledge Discovery and Data Mining V. 2 , pages=

  6. [14]

    Proceedings of the 30th ACM SIGKDD Conference on Knowledge Discovery and Data Mining , pages=

    Cost-efficient fraud risk optimization with submodularity in insurance claim , author=. Proceedings of the 30th ACM SIGKDD Conference on Knowledge Discovery and Data Mining , pages=

  7. [15]

    Internet of Things—Applications and Future: Proceedings of ITAF 2019 , pages=

    Survey of machine learning approaches of anti-money laundering techniques to counter terrorism finance , author=. Internet of Things—Applications and Future: Proceedings of ITAF 2019 , pages=. 2020 , publisher=

  8. [16]

    Knowledge and Information Systems , volume=

    Machine learning techniques for anti-money laundering (AML) solutions in suspicious transaction detection: a review , author=. Knowledge and Information Systems , volume=. 2018 , publisher=

  9. [17]

    Bronstein , title =

    Jake Topping and Francesco Di Giovanni and Benjamin Paul Chamberlain and Xiaowen Dong and Michael M. Bronstein , title =. The Tenth International Conference on Learning Representations,. 2022 , url =

  10. [18]

    Advances in Neural Information Processing Systems , volume=

    Recipe for a general, powerful, scalable graph transformer , author=. Advances in Neural Information Processing Systems , volume=

  11. [19]

    9th International Conference on Learning Representations,

    Uri Alon and Eran Yahav , title =. 9th International Conference on Learning Representations,. 2021 , url =

  12. [20]

    arXiv preprint arXiv:2105.14491 , year=

    How attentive are graph attention networks? , author=. arXiv preprint arXiv:2105.14491 , year=

  13. [21]

    2011 , note =

    Estimating Illicit Financial Flows Resulting from Drug Trafficking and Other Transnational Organized Crimes , institution =. 2011 , note =

  14. [22]

    Lannoo, Karel and Parlour, Richard , title=

  15. [23]

    Graph Feature Preprocessor: Real-time Subgraph-based Feature Extraction for Financial Crime Detection , year =

    Blanu. Graph Feature Preprocessor: Real-time Subgraph-based Feature Extraction for Financial Crime Detection , year =. doi:10.1145/3677052.3698674 , booktitle =

  16. [24]

    Thirty-seventh Conference on Neural Information Processing Systems Datasets and Benchmarks Track , year=

    Realistic Synthetic Financial Transactions for Anti-Money Laundering Models , author=. Thirty-seventh Conference on Neural Information Processing Systems Datasets and Benchmarks Track , year=

  17. [25]

    Proceedings of the Twenty-Ninth International Joint Conference on Artificial Intelligence , articleno =

    Zhang, Shuo and Xie, Lei , title =. Proceedings of the Twenty-Ninth International Joint Conference on Artificial Intelligence , articleno =. 2021 , isbn =

  18. [26]

    ACM Comput

    Chandola, Varun and Banerjee, Arindam and Kumar, Vipin , title =. ACM Comput. Surv. , month = jul, articleno =. 2009 , issue_date =. doi:10.1145/1541880.1541882 , abstract =

  19. [27]

    ANFIS based sensor fault detection for continuous stirred tank reactor , journal =

    U. ANFIS based sensor fault detection for continuous stirred tank reactor , journal =. 2011 , note =. doi:https://doi.org/10.1016/j.asoc.2010.10.009 , url =

  20. [28]

    Spam bot detection on Twitter platform using Positional Attention based Dense Convolutional Neural Network , journal =

    Hemal Girishkumar Shah and Hiren Joshi , keywords =. Spam bot detection on Twitter platform using Positional Attention based Dense Convolutional Neural Network , journal =. 2025 , issn =. doi:https://doi.org/10.1016/j.asoc.2025.113725 , url =

  21. [29]

    and Yo-Ping Huang , booktitle=

    Yufeng Kou and Chang-Tien Lu and Sirwongwattana, S. and Yo-Ping Huang , booktitle=. Survey of fraud detection techniques , year=

  22. [30]

    Principal Neighbourhood Aggregation for Graph Nets , booktitle =

    Gabriele Corso and Luca Cavalleri and Dominique Beaini and Pietro Li. Principal Neighbourhood Aggregation for Graph Nets , booktitle =. 2020 , url =

  23. [31]

    7th International Conference on Learning Representations,

    Keyulu Xu and Weihua Hu and Jure Leskovec and Stefanie Jegelka , title =. 7th International Conference on Learning Representations,. 2019 , url =

  24. [32]

    Graph Attention Networks , booktitle =

    Petar Velickovic and Guillem Cucurull and Arantxa Casanova and Adriana Romero and Pietro Li. Graph Attention Networks , booktitle =. 2018 , url =

  25. [33]

    IEEE transactions on neural networks and learning systems , volume=

    Credit card fraud detection: a realistic modeling and a novel learning strategy , author=. IEEE transactions on neural networks and learning systems , volume=. 2017 , publisher=

  26. [34]

    Big Data and Cognitive Computing , volume=

    Enhancing credit card fraud detection: an ensemble machine learning approach , author=. Big Data and Cognitive Computing , volume=. 2024 , publisher=

  27. [35]

    Applied Soft Computing , volume=

    Value-added tax fraud detection with scalable anomaly detection techniques , author=. Applied Soft Computing , volume=. 2020 , publisher=

  28. [36]

    Mathematical Problems in Engineering , volume=

    Towards tax evasion detection using improved particle swarm optimization algorithm , author=. Mathematical Problems in Engineering , volume=. 2022 , publisher=

  29. [38]

    Ecology , volume=

    Diversity and evenness: a unifying notation and its consequences , author=. Ecology , volume=. 1973 , publisher=

  30. [39]

    Oikos , volume=

    Entropy and diversity , author=. Oikos , volume=. 2006 , publisher=

  31. [40]

    The Bell system technical journal , volume=

    A mathematical theory of communication , author=. The Bell system technical journal , volume=. 1948 , publisher=

  32. [41]

    Masked Label Prediction: Unified Message Passing Model for Semi-Supervised Classification , booktitle =

    Yunsheng Shi and Zhengjie Huang and Shikun Feng and Hui Zhong and Wenjing Wang and Yu Sun , editor =. Masked Label Prediction: Unified Message Passing Model for Semi-Supervised Classification , booktitle =. 2021 , url =. doi:10.24963/IJCAI.2021/214 , timestamp =

  33. [43]

    Proceedings of the 5th ACM International Conference on AI in Finance , pages=

    FraudGT: a simple, effective, and efficient graph transformer for financial fraud detection , author=. Proceedings of the 5th ACM International Conference on AI in Finance , pages=

  34. [44]

    Advances in neural information processing systems , volume=

    Inductive representation learning on large graphs , author=. Advances in neural information processing systems , volume=

  35. [45]

    Kipf and Max Welling , title =

    Thomas N. Kipf and Max Welling , title =. 5th International Conference on Learning Representations,. 2017 , url =

  36. [46]

    Advances in Neural Information Processing Systems , volume=

    Generalised f-mean aggregation for graph neural networks , author=. Advances in Neural Information Processing Systems , volume=

  37. [47]

    Neural Networks , volume=

    BalancerGNN: Balancer Graph Neural Networks for imbalanced datasets: A case study on fraud detection , author=. Neural Networks , volume=. 2025 , publisher=

  38. [48]

    IEEE Access , volume=

    Scalable semi-supervised graph learning techniques for anti money laundering , author=. IEEE Access , volume=. 2024 , publisher=

  39. [49]

    Big Data and Cognitive Computing , VOLUME =

    Khalid, Abdul Rehman and Owoh, Nsikak and Uthmani, Omair and Ashawa, Moses and Osamor, Jude and Adejoh, John , TITLE =. Big Data and Cognitive Computing , VOLUME =. 2024 , NUMBER =

  40. [50]

    Ensemble of deep sequential models for credit card fraud detection , journal =

    Javad Forough and Saeedeh Momtazi , keywords =. Ensemble of deep sequential models for credit card fraud detection , journal =. 2021 , issn =. doi:https://doi.org/10.1016/j.asoc.2020.106883 , url =

  41. [51]

    Cost Sensitive Credit Card Fraud Detection Using Bayes Minimum Risk , year=

    Bahnsen, Alejandro Correa and Stojanovic, Aleksandar and Aouada, Djamila and Ottersten, Björn , booktitle=. Cost Sensitive Credit Card Fraud Detection Using Bayes Minimum Risk , year=

  42. [52]

    Decision support systems , volume=

    APATE: A novel approach for automated credit card transaction fraud detection using network-based extensions , author=. Decision support systems , volume=. 2015 , publisher=

  43. [53]

    Expert systems with applications , volume=

    Learned lessons in credit card fraud detection from a practitioner perspective , author=. Expert systems with applications , volume=. 2014 , publisher=

  44. [54]

    Proceedings of the AAAI Conference on Artificial Intelligence , volume=

    Provably powerful graph neural networks for directed multigraphs , author=. Proceedings of the AAAI Conference on Artificial Intelligence , volume=

  45. [55]

    Advances in Neural Information Processing Systems , volume=

    Approximation ratios of graph neural networks for combinatorial problems , author=. Advances in Neural Information Processing Systems , volume=

  46. [56]

    Proceedings of the web conference 2021 , pages=

    Pick and choose: a GNN-based imbalanced learning approach for fraud detection , author=. Proceedings of the web conference 2021 , pages=

  47. [57]

    Proceedings of the AAAI conference on artificial intelligence , volume=

    Spatio-temporal attention-based neural network for credit card fraud detection , author=. Proceedings of the AAAI conference on artificial intelligence , volume=

  48. [58]

    Algorithms , volume=

    A Temporal Graph Network Algorithm for Detecting Fraudulent Transactions on Online Payment Platforms , author=. Algorithms , volume=. 2024 , publisher=

  49. [59]

    Computers & Security , volume=

    Reducing false positives in bank anti-fraud systems based on rule induction in distributed tree-based models , author=. Computers & Security , volume=. 2022 , publisher=

  50. [60]

    International Journal of Accounting Information Systems , volume=

    Reducing false positives in fraud detection: Combining the red flag approach with process mining , author=. International Journal of Accounting Information Systems , volume=. 2018 , publisher=

  51. [61]

    Ieee Access , volume=

    Financial cybercrime: A comprehensive survey of deep learning approaches to tackle the evolving financial crime landscape , author=. Ieee Access , volume=. 2021 , publisher=

  52. [62]

    Procedia Computer Science , volume=

    Unbalanced credit card fraud detection data: A machine learning-oriented comparative study of balancing techniques , author=. Procedia Computer Science , volume=. 2023 , publisher=

  53. [63]

    Fraud detection system: A survey , journal =

    Aisha Abdallah and Mohd Aizaini Maarof and Anazida Zainal , keywords =. Fraud detection system: A survey , journal =. 2016 , issn =. doi:https://doi.org/10.1016/j.jnca.2016.04.007 , url =

  54. [64]

    A Comparative Study of Machine Learning Techniques for Credit Card Fraud Detection Based on Time Variance , year=

    Rajora, Shantanu and Li, Dong-Lin and Jha, Chandan and Bharill, Neha and Patel, Om Prakash and Joshi, Sudhanshu and Puthal, Deepak and Prasad, Mukesh , booktitle=. A Comparative Study of Machine Learning Techniques for Credit Card Fraud Detection Based on Time Variance , year=

  55. [65]

    Proceedings of the 2020 5th International Conference on Machine Learning Technologies , pages =

    Alarab, Ismail and Prakoonwit, Simant and Nacer, Mohamed Ikbal , title =. Proceedings of the 2020 5th International Conference on Machine Learning Technologies , pages =. 2020 , isbn =. doi:10.1145/3409073.3409078 , abstract =

  56. [66]

    arXiv preprint arXiv:2210.07715 , year=

    Not all neighbors are worth attending to: Graph selective attention networks for semi-supervised learning , author=. arXiv preprint arXiv:2210.07715 , year=

  57. [67]

    Sensors , volume=

    Deep learning-based anomaly detection in video surveillance: A survey , author=. Sensors , volume=. 2023 , publisher=

  58. [68]

    Applied Sciences , volume=

    Insider threat detection based on user behavior modeling and anomaly detection algorithms , author=. Applied Sciences , volume=. 2019 , publisher=

  59. [69]

    Proceedings of the VLDB Endowment , volume=

    Activeclean: Interactive data cleaning for statistical modeling , author=. Proceedings of the VLDB Endowment , volume=. 2016 , publisher=

  60. [70]

    2021 IEEE 17th international conference on intelligent computer communication and processing (ICCP) , pages=

    Human behavior and anomaly detection using machine learning and wearable sensors , author=. 2021 IEEE 17th international conference on intelligent computer communication and processing (ICCP) , pages=. 2021 , organization=

  61. [71]

    Expert systems With applications , volume=

    Financial fraud: a review of anomaly detection techniques and recent advances , author=. Expert systems With applications , volume=. 2022 , publisher=

  62. [72]

    Internet of Things , volume=

    Anomaly detection system for data quality assurance in IoT infrastructures based on machine learning , author=. Internet of Things , volume=. 2024 , publisher=

  63. [73]

    ICASSP 2021-2021 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP) , pages=

    A short tutorial on the weisfeiler-lehman test and its variants , author=. ICASSP 2021-2021 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP) , pages=. 2021 , organization=

  64. [74]

    European semantic web conference , pages=

    Modeling relational data with graph convolutional networks , author=. European semantic web conference , pages=. 2018 , organization=

  65. [75]

    Proceedings of the AAAI conference on artificial intelligence , volume=

    Identity-aware graph neural networks , author=. Proceedings of the AAAI conference on artificial intelligence , volume=

  66. [76]

    Finance and Corporate Governance Conference , year=

    A Comparative Analysis of the Extent of Money Laundering in Australia, UAE, UK and the USA , author=. Finance and Corporate Governance Conference , year=

  67. [77]

    Journal of Computational Informatics & Business , volume=

    Enhancing Anti-Money Laundering Systems with Machine Learning: A Comparative Analysis of Supervised Models , author=. Journal of Computational Informatics & Business , volume=

  68. [78]

    Discover Artificial Intelligence , volume=

    Comparative analysis of machine learning algorithms for money laundering detection , author=. Discover Artificial Intelligence , volume=. 2025 , publisher=

  69. [79]

    International conference on machine learning , pages=

    On over-squashing in message passing neural networks: The impact of width, depth, and topology , author=. International conference on machine learning , pages=. 2023 , organization=

  70. [80]

    IEEE Transactions on Knowledge and Data Engineering , volume=

    Sparse graph attention networks , author=. IEEE Transactions on Knowledge and Data Engineering , volume=. 2021 , publisher=

  71. [81]

    Wiley Interdisciplinary Reviews: Computational Statistics , volume=

    Advances in Continual Graph Learning for Anti-Money Laundering Systems: A Comprehensive Review , author=. Wiley Interdisciplinary Reviews: Computational Statistics , volume=. 2025 , publisher=

  72. [82]

    2025 , publisher=

    Chen, Chen and Ma, Mingyang and Nie, Jiani and Wang, Lei and Xu, Jian , journal=. 2025 , publisher=

  73. [83]

    V.; Egressy, B.; Anghel, A.; and Atasu, K

    Altman, E.; Blanu s a, J.; Niederh \"a usern, L. V.; Egressy, B.; Anghel, A.; and Atasu, K. 2023. Realistic Synthetic Financial Transactions for Anti-Money Laundering Models. In Thirty-seventh Conference on Neural Information Processing Systems Datasets and Benchmarks Track

  74. [84]

    D.; and Vorst, P

    Beneish, M. D.; and Vorst, P. 2022. The cost of fraud prediction errors. The Accounting Review, 97(6): 91--121

  75. [85]

    C .; Chen, L

    Bilgi, H. C .; Chen, L. Y.; and Atasu, K. 2024. Multigraph Message Passing with Bi-Directional Multi-Edge Aggregations. arXiv preprint arXiv:2412.00241

  76. [86]

    Blanu s a, J.; Cravero Baraja, M.; Anghel, A.; von Niederh\" a usern, L.; Altman, E.; Pozidis, H.; and Atasu, K. 2024. Graph Feature Preprocessor: Real-time Subgraph-based Feature Extraction for Financial Crime Detection. In Proceedings of the 5th ACM International Conference ...

  77. [87]

    Boyapati, M.; and Aygun, R. 2025. BalancerGNN: Balancer Graph Neural Networks for imbalanced datasets: A case study on fraud detection. Neural Networks, 182: 106926

  78. [88]

    Chen, C.; Ma, M.; Nie, J.; Wang, L.; and Xu, J. 2025. PNAGMDA : A Principal Neighborhood Aggregation Based Graph Neural Network for miRNA -Disease Association Prediction. IEEE/ACM Transactions on Computational Biology and Bioinformatics

  79. [89]

    D.; Teoh, E

    Chen, Z.; Van Khoa, L. D.; Teoh, E. N.; Nazir, A.; Karuppiah, E. K.; and Lam, K. S. 2018. Machine learning techniques for anti-money laundering (AML) solutions in suspicious transaction detection: a review. Knowledge and Information Systems, 57(2): 245--285

  80. [90]

    Corso, G.; Cavalleri, L.; Beaini, D.; Li \` o , P.; and Velickovic, P. 2020. Principal Neighbourhood Aggregation for Graph Nets. In Larochelle, H.; Ranzato, M.; Hadsell, R.; Balcan, M.; and Lin, H., eds., Advances in Neural Information Processing Systems 33: Annual Conference ...

  81. [91]

    Di Giovanni, F.; Giusti, L.; Barbero, F.; Luise, G.; Lio, P.; and Bronstein, M. M. 2023. On over-squashing in message passing neural networks: The impact of width, depth, and topology. In International conference on machine learning, 7865--7885. PMLR

  82. [92]

    Egressy, B.; Von Niederh \"a usern, L.; Blanu s a, J.; Altman, E.; Wattenhofer, R.; and Atasu, K. 2024. Provably powerful graph neural networks for directed multigraphs. In Proceedings of the AAAI Conference on Artificial Intelligence, volume 38, 11838--11846

  83. [93]

    Gounoue, S.; Sao, A.; and Gottschalk, S. 2025. TeMP-TraG: Edge-based Temporal Message Passing in Transaction Graphs. arXiv preprint arXiv:2503.16901

  84. [94]

    A.; Beutel, A.; Shah, N.; and Faloutsos, C

    Hooi, B.; Shin, K.; Song, H. A.; Beutel, A.; Shah, N.; and Faloutsos, C. 2017. Graph-based fraud detection in the face of camouflage. ACM Transactions on Knowledge Discovery from Data (TKDD), 11(4): 1--26

  85. [95]

    T.; and Villar, S

    Huang, N. T.; and Villar, S. 2021. A short tutorial on the weisfeiler-lehman test and its variants. In ICASSP 2021-2021 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), 8533--8537. IEEE

  86. [96]

    R.; Hermsen, F.; Chala, S

    Karim, M. R.; Hermsen, F.; Chala, S. A.; De Perthuis, P.; and Mandal, A. 2024. Scalable semi-supervised graph learning techniques for anti money laundering. IEEE Access, 12: 50012--50029

  87. [97]

    N.; and Welling, M

    Kipf, T. N.; and Welling, M. 2017. 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 . OpenReview.net

  88. [98]

    Kou, Y.; Lu, C.-T.; Sirwongwattana, S.; and Huang, Y.-P. 2004. Survey of fraud detection techniques. In IEEE International Conference on Networking, Sensing and Control, 2004, volume 2, 749--754 Vol.2

  89. [99]

    Lannoo, K.; and Parlour, R. 2021. Anti-Money Laundering in the EU: Time to get serious. Technical report, Centre for European Policy Studies

  90. [100]

    Liang, S.; Xie, Q.; Feng, J.; Yue, Y.; Li, H.; Ye, J.; and Jie, W. 2025. Feature Reconstruction for Anomaly Detection on Directed Multigraphs: A Preprocessing Framework for GNNs. In Proceedings of the 31st ACM SIGKDD Conference on Knowledge Discovery and Data Mining V. 2, 1707--1717

  91. [101]

    Lin, J.; Guo, X.; Zhu, Y.; Mitchell, S.; Altman, E.; and Shun, J. 2024. FraudGT: a simple, effective, and efficient graph transformer for financial fraud detection. In Proceedings of the 5th ACM International Conference on AI in Finance, 292--300

  92. [102]

    Liu, S.; Hooi, B.; and Faloutsos, C. 2018. A contrast metric for fraud detection in rich graphs. IEEE Transactions on Knowledge and Data Engineering, 31(12): 2235--2248

  93. [103]

    Liu, Y.; Ao, X.; Qin, Z.; Chi, J.; Feng, J.; Yang, H.; and He, Q. 2021. Pick and choose: a GNN-based imbalanced learning approach for fraud detection. In Proceedings of the web conference 2021, 3168--3177

  94. [104]

    Paoletti, G.; Giobergia, F.; Giordano, D.; Cagliero, L.; Ronchiadin, S.; Moncalvo, D.; Mellia, M.; and Baralis, E. 2025. MAD: Multicriteria Anomaly Detection of Suspicious Financial Accounts from Billions of Cash Transactions. In Proceedings of the 31st ACM SIGKDD Conference o...

  95. [105]

    Rees, B.; Wang, X.; Eaton, J.; Yilmaz, O.; Ratzel, R.; and LaSalle, D. 2022. Accelerated GNN training with DGL and RAPIDS cuGraph in a Fraud Detection Workflow. In Proceedings of the 28th ACM SIGKDD Conference on Knowledge Discovery and Data Mining, 4820--4821

  96. [106]

    Salda \ n a-Ulloa, D.; De Ita Luna, G.; and Marcial-Romero, J. R. 2024. A Temporal Graph Network Algorithm for Detecting Fraudulent Transactions on Online Payment Platforms. Algorithms, 17(12): 552

  97. [107]

    Sato, R.; Yamada, M.; and Kashima, H. 2019. Approximation ratios of graph neural networks for combinatorial problems. Advances in Neural Information Processing Systems, 32

  98. [108]

    N.; Bloem, P.; Van Den Berg, R.; Titov, I.; and Welling, M

    Schlichtkrull, M.; Kipf, T. N.; Bloem, P.; Van Den Berg, R.; Titov, I.; and Welling, M. 2018. Modeling relational data with graph convolutional networks. In European semantic web conference, 593--607. Springer

  99. [109]

    Schmidt, J.; Pasadakis, D.; Sathe, M.; and Schenk, O. 2024. GAMLNet: a graph based framework for the detection of money laundering. In 2024 11th IEEE Swiss Conference on Data Science (SDS), 241--245. IEEE

  100. [110]

    Shi, Y.; Huang, Z.; Feng, S.; Zhong, H.; Wang, W.; and Sun, Y. 2021. Masked Label Prediction: Unified Message Passing Model for Semi-Supervised Classification. In Zhou, Z., ed., Proceedings of the Thirtieth International Joint Conference on Artificial Intelligence, IJCAI 2021,...

  101. [111]

    Sun, J.; Chang, W.; Yang, C.; Huang, Q.; and Wang, S. 2026. Relieving the Over-Aggregating Effect in Graph Transformers. Advances in Neural Information Processing Systems, 38: 6673--6699

  102. [112]

    Suzumura, T.; and Kanezashi, H. 2021. Anti-Money Laundering Datasets : InPlusLab Anti-Money Laundering DataDatasets. http://github.com/IBM/AMLSim/

  103. [113]

    United Nations Office on Drugs and Crime . 2011. Estimating Illicit Financial Flows Resulting from Drug Trafficking and Other Transnational Organized Crimes. Technical report, UNODC. Global money laundering estimates: 2-5\

  104. [114]

    Velickovic, P.; Cucurull, G.; Casanova, A.; Romero, A.; Li \` o , P.; and Bengio, Y. 2018. Graph Attention Networks. In 6th International Conference on Learning Representations, ICLR 2018, Vancouver, BC, Canada, April 30 - May 3, 2018, Conference Track Proceedings . OpenReview.net

  105. [115]

    Wu, J.; He, J.; and Xu, J. 2019. Net: Degree-specific graph neural networks for node and graph classification. In Proceedings of the 25th ACM SIGKDD international conference on knowledge discovery & data mining, 406--415

  106. [116]

    P.; and Yan, J

    Wu, Q.; Zhao, W.; Li, Z.; Wipf, D. P.; and Yan, J. 2022. Nodeformer: A scalable graph structure learning transformer for node classification. Advances in neural information processing systems, 35: 27387--27401

  107. [117]

    Xu, K.; Hu, W.; Leskovec, J.; and Jegelka, S. 2019. How Powerful are Graph Neural Networks? In 7th International Conference on Learning Representations, ICLR 2019, New Orleans, LA, USA, May 6-9, 2019 . OpenReview.net

  108. [118]

    Yang, C.; Liu, H.; Wang, D.; Zhang, Z.; Yang, C.; and Shi, C. 2025. Flag: Fraud detection with llm-enhanced graph neural network. In Proceedings of the 31st ACM SIGKDD Conference on Knowledge Discovery and Data Mining V. 2, 5150--5160

  109. [119]

    M.; Ying, R.; and Leskovec, J

    You, J.; Gomes-Selman, J. M.; Ying, R.; and Leskovec, J. 2021. Identity-aware graph neural networks. In Proceedings of the AAAI conference on artificial intelligence, volume 35, 10737--10745

  110. [120]

    Zhang, S.; and Xie, L. 2021. Improving attention mechanism in graph neural networks via cardinality preservation. In Proceedings of the Twenty-Ninth International Joint Conference on Artificial Intelligence, IJCAI'20. ISBN 9780999241165

Pith tools

Reviewed July 14, 2026 · model on record in the stance chip above.