A Fully Bayesian Approach for Generating Synthetic Marks and Geographies for Confidential Data

• Activity Number: 18
• Type: Topic Contributed
• Date/Time: Sunday, August 3, 2014 : 2:00 PM to 3:50 PM
• Sponsor: Survey Research Methods Section
• Author(s): Harrison Quick*+ and Scott Holan and Christopher K. Wikle and Jerome P. Reiter
• Companies: University of Missouri and University of Missouri and University of Missouri and Duke University
• Keywords: Point process ; Spatial ; Survey data ; Disclosure ; Predictive Process

Abstract

Due to the vast amount of information available to the public, releasing data with fine geographies, while protecting the confidentiality of data subjects' identities and attributes, can be a challenging endeavor. Oftentimes, data stewards resort to aggregating subjects into small areal regions, such as census tracts, or swapping sensitive attributes between subjects. Though such measures have been shown to reduce data privacy risks, this comes at the expense of data quality. Our goal is to protect the privacy of the data subjects at fine scales of geography without compromising the validity of the resulting inference. To accomplish this goal, we propose a fully Bayesian approach for generating synthetic data that maintain both the statistical properties and spatial dependence structure of the original data. We illustrate our approach through simulation as well as a real-data example.