Disclosure Risks for Microdata

Abstract

In this chapter we consider the potential disclosure risk arising from the release of microdata. We suppose that the data consist of a standard rectangular data file containing values of variables which at this stage have undergone no SDC treatment. We consider first possible scenarios by which an intruder might attempt to achieve disclosure. This enables us to specify a framework within which disclosure risk may be defined in terms of an intruder’s predictive probability distribution for values of confidential variables. Following a discussion of this predictive approach to measuring disclosure risk, we present arguments for preferring instead to measure risk in terms of the probability of re-identification. The estimation of re-identification risk is discussed in general and for the important special case of discrete variables.

I have the will nor the right to blame

Yet to many (though not to all)

The sweets of destruction are somewhat tame

When no personal risks befall;

—A.L. GORDON, Ye Wearie Wayfarer, hys Ballad in Eight Fyttes.