Developing and Testing the Microdata Analysis System

• Activity Number: 411
• Type: Topic Contributed
• Date/Time: Tuesday, August 11, 2015 : 2:00 PM to 3:50 PM
• Sponsor: Government Statistics Section
• Author(s): Michael Freiman* and Amy Lauger and Marlow Lemons and Bryan Schar
• Companies: U.S. Census Bureau and U.S. Census Bureau and U.S. Census Bureau and U.S. Census Bureau
• Keywords: confidentiality ; disclosure avoidance ; remote access ; tabulation

Abstract

The Census Bureau is developing a Microdata Analysis System (MAS) that will provide custom tables and other analyses based on user specifications. The underlying microdata will be fully protected because calculations for the MAS will take place behind two firewalls and only the final output will be provided to the user. Results will be given only if the output passes a set of disclosure rules. We describe the planned capabilities of the MAS, as well as the testing and evaluation of the disclosure protection measures and some attacks that are being evaluated to ensure that the system only gives data that are safe to release. We focus on assessing and preventing one type of attack: a differencing attack in which the user produces the same table for two very similar universes and subtracts the numbers in the two tables to reveal information about an individual respondent. We consider the prevalence and risks of such an attack in the absence of protections and how the disclosure measures in place protect against an attack.