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.

Differential Privacy With Variant-Noise For Gaussian Processes Classification

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

Differentially Private Regression and Classification with Sparse Gaussian Processes

A New Noise Generating Method Based on Gaussian Sampling for Privacy Preservation

Differential Privacy of Noisy (S)GD under Heavy-Tailed Perturbations

Securely Sampling Discrete Gaussian Noise for Multi-Party Differential Privacy

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

Differentially Private Variational Inference for Non-conjugate Models

Noise-Aware Differentially Private Variational Inference

Noise-Aware Differentially Private Regression via Meta-Learning

Generalized Gaussian Mechanism for Differential Privacy

Noise Variance Optimization in Differential Privacy: A Game-Theoretic Approach Through Per-Instance Differential Privacy

The Discrete Gaussian for Differential Privacy

On the Privacy of Selection Mechanisms with Gaussian Noise

Improved Matrix Gaussian Mechanism for Differential Privacy

Deep Learning with Gaussian Differential Privacy

Differentially private stochastic gradient descent with low-noise

Reviewing and Improving the Gaussian Mechanism for Differential Privacy

Less is More: Revisiting the Gaussian Mechanism for Differential Privacy

Differential Privacy Mechanisms in Neural Tangent Kernel Regression

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