DP-perturbed GNN explanations leak graph structure, enabling reconstruction via reverse diffusion with AUC above 0.7 at epsilon=5 on five of seven benchmarks.
Interpretability and privacy in machine learn- ing: Studying the trade-off through attribution methods
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Graph Reconstruction from Differentially Private GNN Explanations
DP-perturbed GNN explanations leak graph structure, enabling reconstruction via reverse diffusion with AUC above 0.7 at epsilon=5 on five of seven benchmarks.