English

The Optimal Mechanism in Differential Privacy: Multidimensional Setting

Cryptography and Security 2013-12-04 v1

Abstract

We derive the optimal ϵ\epsilon-differentially private mechanism for a general two-dimensional real-valued (histogram-like) query function under a utility-maximization (or cost-minimization) framework for the 1\ell^1 cost function. We show that the optimal noise probability distribution has a correlated multidimensional staircase-shaped probability density function. Compared with the Laplacian mechanism, we show that in the high privacy regime (as ϵ0\epsilon \to 0), the Laplacian mechanism is approximately optimal; and in the low privacy regime (as ϵ+\epsilon \to +\infty), the optimal cost is Θ(eϵ3)\Theta(e^{-\frac{\epsilon}{3}}), while the cost of the Laplacian mechanism is 2Δϵ\frac{2\Delta}{\epsilon}, where Δ\Delta is the sensitivity of the query function. We conclude that the gain is more pronounced in the low privacy regime. We conjecture that the optimality of the staircase mechanism holds for vector-valued (histogram-like) query functions with arbitrary dimension, and holds for many other classes of cost functions as well.

Keywords

Cite

@article{arxiv.1312.0655,
  title  = {The Optimal Mechanism in Differential Privacy: Multidimensional Setting},
  author = {Quan Geng and Pramod Viswanath},
  journal= {arXiv preprint arXiv:1312.0655},
  year   = {2013}
}

Comments

18 pages, 2 figures. arXiv admin note: text overlap with arXiv:1212.1186

R2 v1 2026-06-22T02:19:23.475Z