Identifying Money Laundering Risks in Digital Asset Transactions Based on AI Algorithms

This paper explores methods for detecting suspicious cryptocurrency transactions associated with money laundering, leveraging advanced AI algorithms. The study introduces a multi-model framework combining Generative Adversarial Networks (GANs), LSTM, Autoencoder-Based Anomaly Detection Model (ABAD), and other algorithms to address challenges like sample imbalance and noisy data. Graph-based feature engineering and embedding methods are utilized to construct transaction information graphs and extract meaningful patterns. The results demonstrate that the ensemble learning approach significantly outperforms individual models and traditional rule-based systems in detecting suspicious transactions. Despite its success, challenges such as imbalanced datasets, noise, and limited relational features remain. Future research is suggested to enhance model performance through graph neural networks and complex network-based methods. This work underscores the scalability and adaptability of machine learning models for addressing the evolving complexities of cryptocurrency money laundering.