Abstract: Differential privacy provides a rigorous framework to quantify data privacy, and has received considerable interest recently. A randomized mechanism satisfying $(\epsilon, \delta)$-differential privacy (DP) roughly means that, except with a small probability $\delta$, altering a record in a dataset cannot change the probability that an output is seen by more than a multiplicative factor $e^{\epsilon} $. A well-known solution to $(\epsilon, \delta)$-DP is the Gaussian mechanism initiated by Dwork et al. [1] in 2006 with an improvement by Dwork and Roth [2] in 2014, where a Gaussian noise amount $\sqrt{2\ln \frac{2}{\delta}} \times \frac{\Delta}{\epsilon}$ of [1] or $\sqrt{2\ln \frac{1.25}{\delta}} \times \frac{\Delta}{\epsilon}$ of [2] is added independently to each dimension of the query result, for a query with $\ell_2$-sensitivity $\Delta$. Although both classical Gaussian mechanisms [1,2] assume $0 < \epsilon \leq 1$, our review finds that many studies in the literature have used the classical Gaussian mechanisms under values of $\epsilon$ and $\delta$ where the added noise amounts of [1,2] do not achieve $(\epsilon,\delta)$-DP. We obtain such result by analyzing the optimal noise amount $\sigma_{DP-OPT}$ for $(\epsilon,\delta)$-DP and identifying $\epsilon$ and $\delta$ where the noise amounts of classical mechanisms are even less than $\sigma_{DP-OPT}$. Since $\sigma_{DP-OPT}$ has no closed-form expression and needs to be approximated in an iterative manner, we propose Gaussian mechanisms by deriving closed-form upper bounds for $\sigma_{DP-OPT}$. Our mechanisms achieve $(\epsilon,\delta)$-DP for any $\epsilon$, while the classical mechanisms [1,2] do not achieve $(\epsilon,\delta)$-DP for large $\epsilon$ given $\delta$. Moreover, the utilities of our mechanisms improve those of [1,2] and are close to that of the optimal yet more computationally expensive Gaussian mechanism.

Wishart Mechanism for Differentially Private Principal Components Analysis.

Differential Privacy with Higher Utility by Exploiting Coordinate-wise Disparity: Laplace Mechanism Can Beat Gaussian in High Dimensions

Less is More: Revisiting the Gaussian Mechanism for Differential Privacy

A Differential Privacy Mechanism Design Under Matrix-Valued Query

Optimal Differentially Private PCA and Estimation for Spiked Covariance Matrices

Differentially Private Low-dimensional Synthetic Data from High-dimensional Datasets

The optimal mechanism in differential privacy

Differential Privacy via Distributionally Robust Optimization

Introducing the Huber mechanism for differentially private low-rank matrix completion

Perturb-and-Project: Differentially Private Similarities and Marginals

Wasserstein Differential Privacy

Differential Privacy for Class-based Data: A Practical Gaussian Mechanism

Utility Preserving Secure Private Data Release

Distributed Differentially-Private Algorithms for Matrix and Tensor Factorization

Heterogeneous Gaussian Mechanism: Preserving Differential Privacy in Deep Learning with Provable Robustness

A differentially private matrix factorization based on vector perturbation for recommender system

Universal Exact Compression of Differentially Private Mechanisms

On the Noise-Information Separation of a Private Principal Component Analysis Scheme

General Gaussian Noise Mechanisms and Their Optimality for Unbiased Mean Estimation

Reviewing and Improving the Gaussian Mechanism for Differential Privacy

Count on Your Elders: Laplace vs Gaussian Noise