LM-DP-SGD estimates layer-specific MIA risks from shadow models and reweights gradients to give stronger protection to vulnerable layers, improving the privacy-utility trade-off over uniform DP-SGD.
Exploring the limits of differentially private deep learning with group-wise clipping
2 Pith papers cite this work. Polarity classification is still indexing.
2
Pith papers citing it
fields
cs.LG 2years
2026 2representative citing papers
citing papers explorer
-
Mitigating Membership Inference in Intermediate Representations with Differentially Private Training
LM-DP-SGD estimates layer-specific MIA risks from shadow models and reweights gradients to give stronger protection to vulnerable layers, improving the privacy-utility trade-off over uniform DP-SGD.
- Fundamental Limitations of Favorable Privacy-Utility Guarantees for DP-SGD