Edge Perturbation

Edge perturbation, the modification of graph structures by adding or removing edges, is a key technique in both improving and attacking graph neural networks (GNNs). Current research focuses on understanding the dual nature of edge perturbation—its ability to enhance GNN performance through data augmentation and simultaneously compromise it through adversarial attacks—leading to the development of methods for targeted and efficient perturbation. This research is crucial for advancing the robustness and reliability of GNNs across various applications, from social network analysis to drug discovery, by identifying vulnerabilities and developing more resilient models.