Copula Models for Data Disclosure Limitation

• Activity Number: 86
• Type: Contributed
• Date/Time: Sunday, August 4, 2013 : 4:00 PM to 5:50 PM
• Sponsor: Survey Research Methods Section
• Author(s): Mario Trottini*+ and Krish Muralidhar and Rathindra Sarathy
• Companies: University of Alicante (spain) and University of Kentucky and Oklahoma State University
• Keywords: Statistical Disclosure Control ; Copulas ; Conditional Distribution Approach

Abstract

Protecting confidential information in databases from disclosure is an important issue for many organizations. The conditional distribution approach is a recently proposed method for data perturbation that requires the perturbed values of the sensitive variables to be generated from the conditional distribution of the sensitive variables given the nonconfidential variables. It provides high data utility and low disclosure risk but has limited applicability since it requires the knowledge of the joint distribution of the original data, which is not available even in the simplest problems. In recent years researchers have realized that copulas might provide a very useful tool to overcome these problems. Gaussian and t-copulas models have been used to implement the conditional distribution approach and provide safe data which preserve important features of the original data such as marginal distributions, rank order correlation and tail dependence. The application of copula based models for statistical disclosure limitation is still in its infancy. In this paper we review the state of the art and briefly outline the many potential applications that still remain to be investigated.