A Joint Exponential Mechanism for Differentially Private Top-k SetDownload PDF

Published: 04 Nov 2021, Last Modified: 15 May 2023PRIML 2021 PosterReaders: Everyone
Keywords: differential privacy, top-k selection
TL;DR: New algorithm for the DP top-k selection problem based on the joint exponential mechanism
Abstract: We present a novel differentially private algorithm for releasing the set of k elements with the highest counts from a data domain of d elements. We define a ``joint'' instance of the exponential mechanism (EM) whose output space consists of all O(d^k) size-k subsets; yet, we are able to show how to sample from this EM in only time O(dk^3). Experiments suggest that this joint approach can yield utility improvements over the existing state of the art for small problem sizes.
Paper Under Submission: The paper is NOT under submission at NeurIPS
1 Reply