GRAPH NEURAL NETWORKS FOR ANTI-MONEY LAUNDERING IN CROSSBLOCKCHAIN ENVIRONMENTS

DOI: 10.31673/2409-7292.2026.010224

Abstract

The proliferation of cross-chain bridges and Layer-2 aggregation solutions has fundamentally transformed the structure
of blockchain transaction networks, while simultaneously creating new channels for illicit financial activity that evades
traditional anti-money laundering (AML) mechanisms. Early attempts to use data science for blockchain analytics, particularly
clustering and pattern recognition methods on large datasets, laid the necessary foundation for understanding anomalous
behavior in decentralized ledgers – as evidenced by previous work comparing cluster modeling algorithms on big data derived
from distributed ledger records (e.g., comparing cluster model building algorithms on blockchain-derived datasets). However,
these previous methodologies were largely limited to offline analysis and static clustering paradigms. In this paper, we build on
this foundational research by proposing a graph neural network (GNN)-based framework for real-time AML risk assessment
across multiple heterogeneous blockchain networks. In contrast to early cluster-centric approaches, the proposed model builds
a single temporal cross-chain graph that integrates intra-network activity, interactions via cross-chain bridges, and Layer-2
settlement events into a coherent framework. A message-passing GNN architecture is used to learn informative node
embeddings that reflect relational dependencies and temporal dynamics between networks, using time decay mechanisms and
edge type weighting to increase sensitivity to delayed and fragmented money laundering tactics. We derive a probabilistic risk
scoring function based on the trained embeddings and optimize the model using a composite loss function that balances
classification quality, graph smoothness, and temporal stability. The framework is evaluated on real-world data from Ethereum,
BNB Chain, Solana, and TON networks, covering cross-chain transfers, deposits, and P2P transfers. Experimental results
demonstrate a 25–40% reduction in false positives and a 15–18% increase in recall compared to baseline single-network AML
models, while maintaining real-time inference capabilities suitable for industrial use. implementation. Building directly on the
findings of early cluster modeling research and developing them through deep graph representation learning, the proposed
framework offers a scalable, explainable, and technically robust solution for AML monitoring in modern cross-chain blockchain
environments.
Keywords: blockchain, cross-chain, neural networks, distributed resources, information system, monitoring, cluster
modeling.

References
1. Wu, Z., et al. (2021). A comprehensive survey on graph neural networks. IEEE Transactions on Neural
Networks and Learning Systems, 32(1), 4–24. DOI: 10.1109/TNNLS.2020.2978386.
2. Hamilton, W., Ying, R., Leskovec, J. (2017). Inductive representation learning on large graphs. Advances
in Neural Information Processing Systems (NeurIPS). https: // proceedings.neurips.cc/paper/2017/file/5dd9db5e03-
3da9c6fb5ba83c7a7ebea9-Paper.pdf.
3. Weber, M., et al. (2019). Anti-money laundering in Bitcoin: Experimenting with graph convolutional
networks. ACM KDD Workshop on Deep Learning Day.
4. Tao, Bishenghui, Ivan Wang-Hei Ho, and Hong-Ning Dai. Complex network analysis of the bitcoin
blockchain network. In: 2021 IEEE International Symposium on Circuits and Systems (ISCAS). IEEE, 2021. p. 1-5.
https://ira.lib.polyu.edu.hk/bitstream/10397/107099/1/Ho_Complex_Network_Analysis.pdf.
5. Drobyk, O. V., Hashko, A. O. (2025). Integration of hybrid mathematical models for cluster-based risk
detection in multi-network blockchain environments. In: Proceedings of the 1st International Scientific and Practical
Conference “Applied Control Systems and Robotics”, Kyiv, Ukraine, (page# 185).
6. Bondarchuk, A., Hashko, A., et al. (2025). Security challenges of blockchain interoperability mechanisms.
In: Cybersecurity Providing in Information and Telecommunication Systems II (CPITS-II). CEUR Workshop
Proceedings, Vol. 4145. URL: https://ceur-ws.org/Vol-4145/paper17.pdf.
7. Гашко А. О., Стражніков А. А. (2025). Порівняння алгоритмів побудови кластерної моделі на базі
набору даних (dataset), отриманого з bigdata. Зв’язок, (1), 49-54. URL: https: // con.dut.edu.ua / index.php /
communication/article/view/2839/2736.
8. Liu, Chenlei, Yuhua Xu, and Zhixin Sun. "Directed dynamic attribute graph anomaly detection based on
evolved graph attention for blockchain." Knowledge and Information Systems 66.2 (2024): 989-1010. https://www.
researchsquare.com/article/rs-3212327/latest.pdf.
9. Hashko, A. O., & Bondarchuk, A. P. (2025). AUTOMATED METHOD FOR VERIFYING THE
CORRECTNESS OF THE EXECUTION OF SMART CONTRACTS IN THE BLOCKCHAIN NETWORK.
Сучасний захист інформації, (4), 37-43. DOI: 10.31673/2409-7292.2025.041204.
10. Zhang, M., Zhang, X., Zhang, Y., & Lin, Z. (2024, September). Security of cross-chain bridges: Attack
surfaces, defenses, and open problems. In Proceedings of the 27th International Symposium on Research in Attacks,
Intrusions and Defenses (pp. 298-316). https://dl.acm.org/doi/pdf/10.1145/3678890.3678894.
11. Hashko, A. O. (2025). The GRX model as a universal approach to multidimensional clustering and risk
assessment in AML systems for crypto-assets. In: Proceedings of the 1st International Scientific and Practical
Conference “Applied Control Systems and Robotics”, Kyiv, Ukraine, (page# 204).
12. Kou, G., et al. (2021). Evaluation of machine learning methods for financial fraud detection under class
imbalance. Expert Systems with Applications, 180, 115067. DOI: 10.1016/j.eswa.2021.115067.
13. Гашко, А. О., et al (2025). Автоматизований метод перевірки правильності виконання смартконтрактів в блокчейн мережі. Телекомунікаційні та інформаційні технології, (1), 13-20. DOI: 10.31673/2412-
4338.2025.014506.
14. Shantyr, Anton, et al. "Prediction of quality software quality indicators with applied modifications of
integrated gradiates methods." Informatyka, Automatyka, Pomiary w Gospodarce i Ochronie Środowiska 15.2 (2025):
139-146. http://doi.org/10.35784/iapgos.6892.
15. Чичкарьов, Євген, et al. "Метод вибору ознак для системи виявлення вторгнень з використанням
ансамблевого підходу та нечіткої логіки." Електронне фахове наукове видання «Кібербезпека: освіта, наука,
техніка» 1.21 (2023): 234-251. https://csecurity.kubg.edu.ua/index.php/journal/article/download/523/409.
16. Zhebka, Viktoriia, et al. "Methods for Predicting Failures in a Smart Home." Digital Economy Concepts and
Technologies Workshop 2024. Vol. 3665. Germany, 2024. https://elibrary.kubg.edu.ua/id/eprint/48728/.
17. Ribeiro, M. T., Singh, S., Guestrin, C. (2016). “Why should I trust you?” Explaining the predictions of any
classifier. Proceedings of the ACM SIGKDD Conference. https://doi.org/10.1145/2939672.2939778.