Activity Number:
|
591
- Synthetic Data and Data Disclosure
|
Type:
|
Contributed
|
Date/Time:
|
Wednesday, August 1, 2018 : 2:00 PM to 3:50 PM
|
Sponsor:
|
Government Statistics Section
|
Abstract #328363
|
Presentation 1
Presentation 2
|
Title:
|
Differentially Private Multiple Synthesis via an Adaptive Multiplicative Weighting Algorithm
|
Author(s):
|
Evercita Eugenio* and Fang Liu
|
Companies:
|
University of Notre Dame and University of Notre Dame
|
Keywords:
|
differential privacy;
multiplicative weights;
data synthesis;
statistical disclosure limitation
|
Abstract:
|
An increased effort for transparency and accountability has led government agencies, businesses, and social media platforms to publicly release more of their data. The simple anonymized methods that have long been employed do not provide a high level of privacy protection, as data intruders can combine their knowledge with public information to identify subjects. The recently developed differentially private data synthesis (DIPS) methods are built upon the concept of differential privacy that provides a strong mathematical privacy guarantee while aiming to maintain the analytical validity and utility of the released sanitized data. We will present an adaptive mechanism that uses multiplicative weighting and variable selection techniques to account for the important variable relationships to generate DIPS data. Simulation and case study results will be presented along with comparisons to some existing DIPS methods.
|
Authors who are presenting talks have a * after their name.