Graph-Based Fraud Detection in Big Data-Driven Financial Systems

Abstract

The rapid expansion of financial transactions in big data-driven environments has led to an increased risk of fraudulent activities, requiring advanced detection mechanisms beyond traditional rule-based and statistical methods. This study explores graph-based fraud detection techniques by leveraging Graph Neural Networks (GNNs), Graph Attention Networks (GATs), and anomaly detection algorithms to enhance the identification of fraudulent transactions in largescale financial systems. The proposed model constructs transaction graphs, capturing the relationships between entities such as users, merchants, and transaction histories, enabling contextaware fraud detection. Experimental results on real-world financial datasets demonstrate that our approach outperforms conventional machine learning techniques by achieving higher detection accuracy, lower false positive rates, and real-time processing capabilities. Furthermore, the integration of self-supervised learning and dynamic graph embeddings improves adaptability to evolving fraud patterns. This research highlights the efficacy of graph-based AI methods in securing financial transactions and mitigating economic risks in modern financial ecosystems.