A Method for Identifying Anomalous Bitcoin Transactions Based on Image-based Representations of Transaction Features

DOI: https://doi.org/10.62517/jbdc.202601310

Author(s)

Zilin Xu

Affiliation(s)

School of Computer Science, Wuhan University, Wuhan, Hubei, China

Abstract

Cryptocurrencies such as Bitcoin are widely used in financial transactions and cross-border payments due to their decentralized and anonymous nature; however, they also provide a covert channel for anomalous transactions such as money laundering and fraud. Addressing the limitations of traditional abnormal transaction detection methods in utilizing transaction network structures and node attribute information, this paper proposes an abnormal transaction detection method based on Node attribute heatmaps.The method first constructs a directed graph using the Elliptic Bitcoin dataset and extracts local k-hop subgraphs centered on labeled transaction nodes. Subsequently, high-dimensional node attribute features are mapped to node attribute heatmaps, transforming the graph-based anomalous transaction detection problem into an image classification task. Finally, a convolutional neural network is employed to perform binary classification on the image samples. To validate the method’s effectiveness, a comparative experiment with a Graph Convolutional Network (GCN) was designed.Experimental results show that the Node attribute heatmap method outperforms both GCN and traditional adjacency matrix image methods in metrics such as Accuracy, Recall, and F1-score, indicating that this method can more fully capture the behavioral characteristics of transaction nodes, thereby enhancing the performance of anomalous transaction detection.

Keywords

Anomalous Transaction Detection; Bitcoin; Graph Representation Learning; Node Attribute Heatmap; Convolutional Neural Network (CNN)

References

[1] Weber, M., Domeniconi, G., Chen, J., Weidele, D. K. I., Bellei, C., Robinson, T., & Leiserson, C. E. (2019). Anti-money laundering in bitcoin: Experimenting with graph convolutional networks for financial forensics. arXiv preprint arXiv:1908.02591. [2] Kipf, T. N., & Welling, M. (2016). Semi-supervised classification with graph convolutional networks. arXiv preprint arXiv:1609.02907. [3] Veličković, P., Cucurull, G., Casanova, A., Romero, A., Lio, P., & Bengio, Y. (2017). Graph attention networks. arXiv preprint arXiv:1710.10903. [4] Xu, K., Hu, W., Leskovec, J., & Jegelka, S. (2018). How powerful are graph neural networks?. arXiv preprint arXiv:1810.00826. [5] Zhou, J., Cui, G., Hu, S., Zhang, Z., Yang, C., Liu, Z., Wang, L., Li, C., & Sun, M. (2020). Graph neural networks: A review of methods and applications. AI Open, 1, 57–81. https://doi.org/10.1016/j.aiopen.2021.01.001 [6] Niepert, M., Ahmed, M., & Kutzkov, K. (2016, June). Learning convolutional neural networks for graphs. In International conference on machine learning (pp. 2014–2023). PMLR. [7] Hegde, K., Magdon-Ismail, M., Ramanathan, R., & Thapa, B. (2018). Network signatures from image representation of adjacency matrices: Deep/transfer learning for subgraph classification. arXiv preprint arXiv:1804.06275.