Causal Machine Learning for Financial Crime Attribution: Uncovering Hidden Cause-and-Effect Relationships Among Fraud, Money Laundering, and Cybersecurity Incidents in Blockchain Ecosystems

Abstract

Blockchain ecosystems have transformed digital finance by enabling decentralized and transparent transactions, yet these same characteristics have facilitated the emergence of complex financial crimes involving fraud, money laundering, and cyber-enabled illicit activities. Existing machine learning approaches have achieved notable success in detecting suspicious transactions, but most remain fundamentally predictive and provide limited understanding of the underlying mechanisms that generate criminal behavior. This limitation constrains the development of effective interventions, regulatory policies, and financial intelligence strategies. This study proposes a causal machine learning framework for financial crime attribution in blockchain ecosystems, emphasizing the discovery of hidden cause-and-effect relationships among fraudulent activities, laundering processes, and cybersecurity incidents. Using the Elliptic Bitcoin transaction network as a case study, the framework integrates predictive modeling, graph-based feature engineering, explainable artificial intelligence, causal discovery techniques, Double Machine Learning, and heterogeneous treatment-effect estimation within a unified analytical pipeline. The methodology combines temporal validation procedures with structural analyses to investigate how network characteristics and behavioral patterns contribute to illicit financial outcomes. The findings demonstrate that strong predictive performance does not necessarily imply causal importance. While conventional machine learning models effectively identify suspicious transactions, causal analyses reveal that several highly predictive variables possess limited independent influence after accounting for network and temporal confounders. Structural properties associated with community organization, transaction dynamics, and connectivity emerge as important components of the broader mechanisms underlying financial crime propagation. The results further highlight the interconnected nature of fraud, money laundering, and cyber-enabled activities within decentralized financial systems. This research contributes to the growing intersection of causal inference and blockchain analytics by explicitly distinguishing predictive explanations from causal explanations and by providing a transparent framework for intervention-oriented financial intelligence. Although the study is constrained by observational data, proxy-based treatments, and platform-specific characteristics, it establishes a practical foundation for developing more interpretable, accountable, and causally informed approaches to combating financial crimes in digital economies.