Detecting Fraud in Financial Networks: A Semi-Supervised GNN Approach with Granger-Causal Explanations

2025-06-25

AI Summary

Financial network fraud detection faces two key challenges—label sparsity and model opacity—that hinder regulatory deployment. To address these, we propose SAGE-FIN, the first framework integrating semi-supervised graph neural networks (GNNs) with Granger causal inference in a unified architecture. Without requiring predefined network structures, SAGE-FIN enables fraud identification under weak supervision and automatically generates regulatory-compliant causal explanation paths. Methodologically, it constructs a heterogeneous graph encoding both node and edge attributes, jointly modeling transaction temporal dynamics and topological dependencies. Technically, a novel causal attention mechanism bridges GNN representations with Granger causality tests, balancing discriminative performance and interpretability. Evaluated on the real-world Elliptic++ dataset, SAGE-FIN significantly outperforms state-of-the-art baselines and produces verifiable, audit-ready causal pathways—establishing a new paradigm for compliant, trustworthy AI in financial regulation.

Technology Category

• Machine Learning: Causal Learning
• Reasoning under Uncertainty: Causality
• Data Mining & Knowledge Management: Graph Mining, Social Network Analysis & Community

Application Category

• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG

Abstract

Fraudulent activity in the financial industry costs billions annually. Detecting fraud, therefore, is an essential yet technically challenging task that requires carefully analyzing large volumes of data. While machine learning (ML) approaches seem like a viable solution, applying them successfully is not so easy due to two main challenges: (1) the sparsely labeled data, which makes the training of such approaches challenging (with inherent labeling costs), and (2) lack of explainability for the flagged items posed by the opacity of ML models, that is often required by business regulations. This article proposes SAGE-FIN, a semi-supervised graph neural network (GNN) based approach with Granger causal explanations for Financial Interaction Networks. SAGE-FIN learns to flag fraudulent items based on weakly labeled (or unlabelled) data points. To adhere to regulatory requirements, the flagged items are explained by highlighting related items in the network using Granger causality. We empirically validate the favorable performance of SAGE-FIN on a real-world dataset, Bipartite Edge-And-Node Attributed financial network (Elliptic++), with Granger-causal explanations for the identified fraudulent items without any prior assumption on the network structure.

Problem

• Detecting fraud in financial networks with sparse labeled data
• Providing explainable fraud detection using Granger causality
• Overcoming ML opacity for regulatory compliance in finance

Innovation

• Semi-supervised GNN for fraud detection
• Granger-causal explanations for transparency
• Works with weakly labeled financial data