AIM: An Adaptive and Iterative Mechanism for Differentially Private Synthetic Data

TL;DR

AIM tackles the challenge of crafting high-utility synthetic data under differential privacy by framing the problem as a workload-adaptive, query-based measurement and synthesis pipeline. It introduces an adaptive, iterative mechanism that selects the most informative measurements relative to the target workload, privately measures them, and then generates synthetic data from the noisy outcomes. The work provides analytic, high-probability per-query error bounds enabling confidence intervals, and demonstrates that AIM consistently outperforms a broad range of existing mechanisms across diverse experimental settings. These contributions offer practical privacy guarantees with improved utility, facilitating more reliable private data releases for analytics and research.

Abstract

We propose AIM, a new algorithm for differentially private synthetic data generation. AIM is a workload-adaptive algorithm within the paradigm of algorithms that first selects a set of queries, then privately measures those queries, and finally generates synthetic data from the noisy measurements. It uses a set of innovative features to iteratively select the most useful measurements, reflecting both their relevance to the workload and their value in approximating the input data. We also provide analytic expressions to bound per-query error with high probability which can be used to construct confidence intervals and inform users about the accuracy of generated data. We show empirically that AIM consistently outperforms a wide variety of existing mechanisms across a variety of experimental settings.

Figures (1)

• Figure 1: An illustration of a Mallard Duck. Picture from Mabel Osgood Wright, Birdcraft, published 1897.