Leveraging Public Data for Practical Private Query Release
arXiv:2102.08598
Abstract
In many statistical problems, incorporating priors can significantly improve performance. However, the use of prior knowledge in differentially private query release has remained underexplored, despite such priors commonly being available in the form of public datasets, such as previous US Census releases. With the goal of releasing statistics about a private dataset, we present PMW^Pub, which -- unlike existing baselines -- leverages public data drawn from a related distribution as prior information. We provide a theoretical analysis and an empirical evaluation on the American Community Survey (ACS) and ADULT datasets, which shows that our method outperforms state-of-the-art methods. Furthermore, PMW^Pub scales well to high-dimensional data domains, where running many existing methods would be computationally infeasible.
arXiv commentICML 2021
References in corpus (9)
- Scalable Private Learning with PATE
- A simple and practical algorithm for differentially private data release
- Optimizing error of high-dimensional statistical queries under differential privacy
- New Oracle-Efficient Algorithms for Private Synthetic Data Release
- Permute-and-Flip: A new mechanism for differentially private selection
- Limits of Private Learning with Access to Public Data
- Private Query Release Assisted by Public Data
- Privately Answering Classification Queries in the Agnostic PAC Model
- Learning from Mixtures of Private and Public Populations
Cited by in corpus (7)
- Differentially Private Fine-tuning of Language Models
- Deep Learning with Label Differential Privacy
- Synthetic Tabular Data: Methods, Attacks and Defenses
- Public Data-Assisted Mirror Descent for Private Model Training
- Turbo: Effective Caching in Differentially-Private Databases
- Relaxed Marginal Consistency for Differentially Private Query Answering
- Prior-Aware Distribution Estimation for Differential Privacy