|
Activity Number:
|
495
- Formal Privacy from a Statistician's Perspective: Dealing with Survey Data and Statistical Inference
|
|
Type:
|
Invited
|
|
Date/Time:
|
Thursday, August 11, 2022 : 8:30 AM to 10:20 AM
|
|
Sponsor:
|
SSC (Statistical Society of Canada)
|
|
Abstract #320567
|
|
|
Title:
|
Analysis of Differentially Private Synthetic Data Sets
|
|
Author(s):
|
Bei Jiang* and Yangdi Jiang and Linglong Kong and Anne-Sophie Charest
|
|
Companies:
|
University of Alberta and University of Alberta and University of Alberta and Université Laval
|
|
Keywords:
|
Differential Privacy;
Bias correction;
Uncertainty ;
Inferential integrity
|
|
Abstract:
|
Differential Privacy (DP) has gained growing attention given its mathematically rigorous guarantee to prevent the worst case scenario of privacy leak. However, new procedures are needed to analyze DP synthetic datasets when Rubin's combining rules for multiple imputation based synthetic data sets are not valid. In this talk, we consider the multivariate regression setting and discuss how to correct for biases induced by introducing the DP mechanism, and provide a proper accounting of uncertainty. Using real datasets, we show how we can obtain approximately the same inference results while still satisfying DP criterion for privacy protection.
|
Authors who are presenting talks have a * after their name.
