Ethereum Fraud Detection via Joint Transaction Language Model and Graph Representation Learning

2024-09-09

arXiv.org

AI Summary

To address escalating fraud threats on Ethereum, existing methods struggle to effectively integrate transaction semantics, attribute similarity, and account topological structure. This paper proposes the first semantic-driven dual-graph collaborative modeling framework: (1) a transaction language model that maps numeric transactions into semantically rich sentences; (2) a dual-graph architecture comprising an attribute-similarity graph and an account-interaction graph; and (3) a semantic–similarity joint attention mechanism, enabling end-to-end co-training of GNNs and sequential models. Evaluated on real-world Ethereum data, our approach achieves an F1-score of 98.7%, substantially outperforming unimodal baselines and effectively detecting emerging fraud patterns—including mixer-based laundering, phishing, and Ponzi schemes. The core contribution lies in the novel unification of transaction semantics, multi-dimensional similarity, and graph-structured relationships—thereby unlocking the synergistic potential of heterogeneous representations for fraud detection.

Technology Category

• Machine Learning: Graph-based Machine Learning
• Data Mining & Knowledge Management: Graph Mining, Social Network Analysis & Community
• Natural Language Processing: Ethics — Bias, Fairness, Transparency & Privacy

Application Category

• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• User Modeling, Personalization and Recommendation: Accountability, Transparency, and Ethics for personalization

Abstract

Ethereum faces growing fraud threats. Current fraud detection methods, whether employing graph neural networks or sequence models, fail to consider the semantic information and similarity patterns within transactions. Moreover, these approaches do not leverage the potential synergistic benefits of combining both types of models. To address these challenges, we propose TLMG4Eth that combines a transaction language model with graph-based methods to capture semantic, similarity, and structural features of transaction data in Ethereum. We first propose a transaction language model that converts numerical transaction data into meaningful transaction sentences, enabling the model to learn explicit transaction semantics. Then, we propose a transaction attribute similarity graph to learn transaction similarity information, enabling us to capture intuitive insights into transaction anomalies. Additionally, we construct an account interaction graph to capture the structural information of the account transaction network. We employ a deep multi-head attention network to fuse transaction semantic and similarity embeddings, and ultimately propose a joint training approach for the multi-head attention network and the account interaction graph to obtain the synergistic benefits of both.

Problem

• Enhancing Ethereum fraud detection
• Combining semantic and graph features
• Improving transaction anomaly identification

Innovation

• Combines transaction language model
• Integrates graph-based representation learning
• Uses multi-head attention network