Oneshot Differentially Private Top-k Selection

Being able to efficiently and accurately select the top-$k$ elements without privacy leakage is an integral component of various data analysis tasks and has gained significant attention. In this paper, we introduce the \textit{oneshot mechanism}, a fast, low-distortion, and differentially private primitive for the top-$k$ problem. Compared with existing approaches in the literature, our algorithm adds Laplace noise to the counts and releases the top-$k$ noisy counts and their estimates in a oneshot fashion, thereby substantially reducing the computational cost while maintaining satisfying utility. Our proof of privacy for this mechanism relies on a novel coupling technique that is of independent theoretical interest. Finally, we apply the oneshot mechanism to multiple hypothesis testing and ranking from pairwise comparisons and thus obtain their differentially private counterparts.