Sample-and-threshold differential privacy: Histograms and applications

Akash Bharadwaj, Graham Cormode

Keywords: histograms, differential privacy, federated analytics

TL;DR: Sampling alone combined with a threshold is sufficient to give differential privacy for the core problem of histograms.

Abstract:

Federated analytics aims to compute accurate statistics from distributed datasets.
A "Differential Privacy" (DP) guarantee is usually desired by the users of the devices storing the data.
In this work, we prove a strong $(\epsilon, \delta)$-DP guarantee for a highly practical sampling-based procedure to derive histograms. We also provide accuracy guarantees and show how to apply the procedure to estimate quantiles and modes.

Paper Under Submission: The paper is NOT under submission at NeurIPS