Joshua Snoke

Claire McKay Bowen, Joshua Snoke

Differentially private synthetic data generation offers a recent solution to release analytically useful data while preserving the privacy of individuals in the data. In order to utilize these algorithms for public policy decisions, policymakers need an accurate understanding of these algorithms' comparative performance. Correspondingly, data practitioners require standard metrics for evaluating the analytic qualities of the synthetic data. In this paper, we present an in-depth evaluation of several differentially private synthetic data algorithms using actual differentially private synthetic data sets created by contestants in the 2018-2019 National Institute of Standards and Technology Public Safety Communications Research (NIST PSCR) Division's ``Differential Privacy Synthetic Data Challenge.'' We offer analyses of these algorithms based on both the accuracy of the data they created and their usability by potential data providers. We frame the methods used in the NIST PSCR data challenge within the broader differentially private synthetic data literature. We implement additional utility metrics, including two of our own, on the differentially private synthetic data and compare mechanism utility on three categories. Our comparative assessment of the differentially private data synthesis methods and the quality metrics shows the relative usefulness, the general strengths and weaknesses, and offers preferred choices of algorithms and metrics. Finally we describe the implications of our evaluation for policymakers seeking to implement differentially private synthetic data algorithms on future data products.

Alexander G. Ororbia, Fridolin Linder, Joshua Snoke

We present a method for generating synthetic versions of Twitter data using neural generative models. The goal is protecting individuals in the source data from stylometric re-identification attacks while still releasing data that carries research value. Specifically, we generate tweet corpora that maintain user-level word distributions by augmenting the neural language models with user-specific components. We compare our approach to two standard text data protection methods: redaction and iterative translation. We evaluate the three methods on measures of risk and utility. We define risk following the stylometric models of re-identification, and we define utility based on two general word distribution measures and two common text analysis research tasks. We find that neural models are able to significantly lower risk over previous methods with little cost to utility. We also demonstrate that the neural models allow data providers to actively control the risk-utility trade-off through model tuning parameters. This work presents promising results for a new tool addressing the problem of privacy for free text and sharing social media data in a way that respects privacy and is ethically responsible.