Masking Procedures for Microdata Disclosure Limitation

Wayne A. Fuller

Abstract:
Masking methods in which error is added to the elements of a data set prior to data release are described. The nature of confidentiality protection provided by these methods, the costs of such procedures to data providers, and the costs of the procedures to data users are investigated. Estimation methods appropriate for data subjected to a random mask are described. Nonlinear functions of the data vector are included in the estimation discussion.

Keywords:
Confidentiality; measurement error; data transformations.