AL-PUF: A Natural Generalization of K-Anonymization and Micro-Aggregation to Improve Analytic Utility

• Activity Number: 689
• Type: Contributed
• Date/Time: Thursday, August 13, 2015 : 10:30 AM to 12:20 PM
• Sponsor: Survey Research Methods Section
• Author(s): Joshua Borton* and Avinash Singh
• Companies: NORC at the University of Chicago and NORC at the University of Chicago
• Keywords: Statistical Disclosure Limitation ; Aggregate Level Analysis ; Micro-Group ; Micro-Mean ; Subsampling ; Unit Level Analysis

Abstract

Aggregate Level Public Use File (AL-PUF) methodology is an SDL method developed at NORC. It relies on low levels of aggregation for producing arbitrary domain-level summary statistics and subsampling to introduce uncertainty. This results in imprecise estimates for very small domains, for disclosure protection, and precise estimates for large domains, promoting analytic utility. This paper demonstrates how AL-PUF can be described as a generalization of the commonly used SDL methods of k-anonymization and micro-aggregation. In AL-PUF, partitioning of data into small aggregates is similar to k-anonymization and computation of summary statistics for each analytic domain within each aggregate, is like micro-aggregation. Subsampling introduces uncertainty in summary statistics and the resulting data is amenable for standard analysis which is not possible for usual PUFs requiring deterministic suppression to satisfy k-anonymization. For AL-PUF, these traditional methods are used as a starting point and then adjusted to address utility issues. These modifications result not only in more utility, but also quantifies risk and utility allowing the balance to be thoughtfully managed.