Global Confidence Degree Based Graph Neural Network for Financial Fraud Detection

TL;DR

This paper tackles financial fraud detection on graph-structured data by introducing Global Confidence Degree (GCD), a global prototype-based measure of node typicality. The authors design GCD-GNN, a graph neural network that leverages iterative prototype extraction, GCD estimation, and two-stream (typical and atypical) aggregation to counter camouflage and exploit global information. They also offer a lightweight variant that preserves strong performance while boosting convergence and speed. Empirical results on two public datasets show that GCD-GNN surpasses state-of-the-art baselines in AUC, F1-Macro, and G-Mean, and ablations confirm the additive benefits of each component. The approach provides faster training/inference and a robust mechanism to incorporate global, prototype-driven signals into GNN-based fraud detection.

Abstract

Graph Neural Networks (GNNs) are widely used in financial fraud detection due to their excellent ability on handling graph-structured financial data and modeling multilayer connections by aggregating information of neighbors. However, these GNN-based methods focus on extracting neighbor-level information but neglect a global perspective. This paper presents the concept and calculation formula of Global Confidence Degree (GCD) and thus designs GCD-based GNN (GCD-GNN) that can address the challenges of camouflage in fraudulent activities and thus can capture more global information. To obtain a precise GCD for each node, we use a multilayer perceptron to transform features and then the new features and the corresponding prototype are used to eliminate unnecessary information. The GCD of a node evaluates the typicality of the node and thus we can leverage GCD to generate attention values for message aggregation. This process is carried out through both the original GCD and its inverse, allowing us to capture both the typical neighbors with high GCD and the atypical ones with low GCD. Extensive experiments on two public datasets demonstrate that GCD-GNN outperforms state-of-the-art baselines, highlighting the effectiveness of GCD. We also design a lightweight GCD-GNN (GCD-GNN$_{light}$) that also outperforms the baselines but is slightly weaker than GCD-GNN on fraud detection performance. However, GCD-GNN$_{light}$ obviously outperforms GCD-GNN on convergence and inference speed.

Figures (9)

• Figure 1: Aggregation pattern. Aggregate from typical and atypical perspectives. Solid lines represent the aggregation of the typical perspective, while dashed lines represent the aggregation of the atypical perspective. The thickness of the line is directly proportional to the weight value of a node.
• Figure 2: Feature embeddings on T-Finance visualization using two different dimensionality reduction techniques. Red color represents fraudulent nodes, while blue represents benign nodes.
• Figure 3: An illustration of the proposed framework.
• Figure 4: Convergence time (log) and AUC of models on T-Finance.
• Figure 5: Distances analyze on T-Finance (left) and FDCompCN (right).

Theorems & Definitions (3)

• Definition 1: Multi-relation Graph
• Definition 2: Prototype
• Definition 3: Global Confidence Degree