Releasing graph neural networks with differential privacy guarantees
Published in Transactions on Machine Learning Research (TMLR), 2023
PrivGnn is a framework for releasing graph neural networks trained with knowledge from private data while preserving privacy. The method combines knowledge distillation, random subsampling, and noisy labeling, and is analyzed in the Renyi differential privacy framework.
Recommended citation: Iyiola E. Olatunji, Thorben Funke, and Megha Khosla. (2023). "Releasing graph neural networks with differential privacy guarantees." Transactions on Machine Learning Research (TMLR).
Download Paper