Evi-VN: Hard Region Guided Virtual Node Evidence Injection for GNN-Based Fraud Detection
Evi-VN is a new framework that detects hard regions—areas where different Graph Neural Networks (GNNs) disagree on fraud detection. It works by using feature-isolated evidence chains to selectively inject information into the graph structure only at likely difficult nodes. This corrects blind spots missed by standard methods, significantly boosting performance across various GNN architectures.