A Novel Graph-Based Framework for Cryptocurrency Fraud Detection
Rong Yang · Advances in economics, business and management research/Advances in Economics, Business and Management Research · 2025
Cryptocurrencies have been widely applied in in-dustries such as finance and physical trade, with Bitcoin and Ethereum becoming the mainstream cryptocurrencies.However, cryptocurrency transactions face numerous security issues, such as money laundering, Ponzi schemes, and high-investment-plan scams.As a product of blockchain, cryptocurrency transactions possess the characteristics of anonymity and immutability.While this anonymity protects the privacy of the parties involved in a transaction, it significantly increases the difficulty for security agencies and government institutions to monitor and regulate these transactions.Although existing methods for fraud detection achieve high accuracy, they often lack interpretability and fail to help security teams identify the entities linked to fraudulent transactions or offer insights into similar fraudulent patterns.To address these issues and improve the interpretability of fraud detection, we propose an innovative graph-based framework.By gathering multi-dimensional data, we can provide a more detailed and holistic view of each transaction record.We develop a heterogeneous graph to model transaction entities, their related transaction records, and transaction flows.Using graph fusion and reasoning techniques, this model aids in analyzing market and entity behaviors and supports heuristic exploration by experts.Finally, a pre-trained Graph Neural Network (GNN) is utilized to quickly pinpoint fraudulent entities and their associated transactions within the graph.