Chinese Researchers Develop AI Framework Detecting Illegal Bitcoin Transactions with 89.4% Accuracy

A new AI model combining DGNN and LLMs achieves 89.4% accuracy in identifying illicit Bitcoin flows, offering regulators a precise tool for combating crypto-related financial crimes.

Woofun AI reports that a team from the People’s Public Security University of China has created an AI framework to detect illegal Bitcoin transactions, achieving an overall accuracy of 89.4%. The system integrates dynamic graph neural networks, memory mechanisms, and large language models to analyze transaction structures and generate risk assessments. Tested on the Elliptic dataset, the model demonstrated an 89.1% accuracy rate for illegal transaction detection and a 64.5% recall rate, outperforming mainstream baselines.

The research, published in the Journal of Intelligence in May, offers an interpretable solution for regulatory authorities. This development coincides with intensified crackdowns on Bitcoin-linked financial crimes in China, including 3,259 prosecutions for money laundering last year and a recent Inner Mongolia case involving nearly 3 billion yuan in illicit funds.