GraphToxin: Reconstructing Full Unlearned Graphs from Graph Unlearning
GraphToxin is a novel full graph reconstruction attack that recovers entire unlearned graphs by exploiting residual traces in Graph Neural Networks (GNNs). It uses three core modules—gradient matching, curvature matching, and feature smoothness—to reconstruct deleted nodes, links, and sensitive content. The study proves existing defenses like Node-DP are largely ineffective against this powerful threat.