Graph-Based Fraud Detection in Open-Loop Gift Cards: Heterogeneous GNNs, Streaming Feature Stores, and Near-Zero-Lag Anomaly Alerts
Jennifer Amebleh, Emmanuel Igba, Onuh Matthew Ijiga
Abstract
Open-loop gift cards, widely used in retail and financial ecosystems, present significant fraud detection challenges due to their anonymous usage, high transaction velocity, and cross-platform interoperability. Traditional rule-based systems and conventional machine learning approaches often struggle to capture the complex, dynamic, and heterogeneous nature of fraudulent activities in such environments. This review explores the integration of graph-based fraud detection frameworks, with a focus on heterogeneous graph neural networks (GNNs), to model multi-relational entities such as customers, merchants, transactions, and cards. By leveraging the structural and semantic richness of heterogeneous graphs, GNNs enable more robust detection of anomalous patterns that may otherwise remain hidden. Furthermore, the paper discusses the role of streaming feature stores for real-time data ingestion, transformation, and storage, ensuring continuous feature updates that support scalable fraud detection pipelines. Special emphasis is placed on near-zero-lag anomaly alerts, which combine low-latency graph analytics with streaming infrastructures to minimize detection delays and reduce financial losses. Through an examination of state-of-the-art techniques, deployment challenges, and emerging opportunities, this review highlights the transformative potential of graph-based AI methods in safeguarding open-loop gift card ecosystems, while outlining future research directions in explainability, scalability, and regulatory compliance.