A convex optimization framework produces DP noise distributions that beat Gaussian and Laplace mechanisms with the same variance in moderate composition regimes, by minimizing Rényi DP at a tuned order.
P., Kosut, O., and Sankar, L
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.IT 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Optimizing Noise Distributions for Differential Privacy
A convex optimization framework produces DP noise distributions that beat Gaussian and Laplace mechanisms with the same variance in moderate composition regimes, by minimizing Rényi DP at a tuned order.