The Investigation of Illicit Tron Wallet based on Temporal Behaviour Analysis
Meng Ling Tsai, Chen Yu Kuan, Chien Lung Lin, Chih Hung Shih, Fu-Ching Tsai · Procedia Computer Science · 2024
The widespread use of cryptocurrency for criminal activities has led to the development of new methods for identifying illicit wallet behavior. This study proposes a machine learning approach that utilizes law enforcement databases to obtain case-related labels on TRON. By analyzing behavioral patterns of wallet addresses in open ledgers, machine learning algorithms were trained to recognize illicit wallet behaviors. The results show that temporal features exhibit improved performance in identifying illicit wallet behavior. The experimental results also indicate that the LightGBM and XGBoost algorithms achieved better performance, while the LightGBM algorithm excelled in Precision and the XGBoost algorithm performed better in Recall. The results of this study indicate that compared to previous research on identifying illegal wallets, criminal behavior develops corresponding criminal patterns based on different blockchain ecosystems. Therefore, for the identification of illegal wallets, it is necessary to extract and analyze features according to different public ledger formats to achieve higher accuracy.