Fraud Syndicates Within Digital Ecosystem: Graph Network and Transaction Analysis Approach

This paper aims to develop and test methods to detect organized crime, fraud syndicates, and Money Laundering schemes within an e-wallet ecosystem. Analytical An analyticalprocess framework combining Graph Analytics and Supervised Learning is developed and trained with sample spaces of fraud and non-fraud. The pipeline utilized Heterogeneous Graph Transformation (HGT), Graph Statistics (Centrality Measures and Community Detection), and a Gradient Boosting Model to produce models for the detection of fraud syndicate syndicatesand organized crime. Welch’s t-test is employed to infer variance differences between samples. Findings confirm the hypothesis that fraud networks are markedly different, exhibiting a more centralized and isolated network compared to the organic, interconnected behaviors of non-fraudulent users. Fraud networks are further characterized by multiple isolated clusters, indicating distinctive groups or behaviors. The proposed method can provide validated methods of fraud and money laundering detection, especially for financial decision-makers and policymakers, to enhance fraud detection systems by improving the protection, integrity, and security of customers and digital transactions.