An attacker can exploit explanation heatmaps from deployed graph neural networks to train a surrogate that matches both predictions and highlighted decision logic, outperforming prior stealing attacks.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Do Explanations Increase the Risk of Decision Logic Leakage? Explanation-Guided Stealing of Graph Models
An attacker can exploit explanation heatmaps from deployed graph neural networks to train a surrogate that matches both predictions and highlighted decision logic, outperforming prior stealing attacks.