Decision Theoretical Based System for Information Downgrading

摘要

It is sometimes necessary for the owner of proprietary data to publicize some of it while keeping the rest as private. For example, when releasing census data or corporate financial information, the release must be conducted in a manner consistent with individual privacy. The process of publicly releasing formerly private data is called downgrading. However, it may be possible to infer unreleased private information from the downgraded public information--the so called inference problem. Here, we discuss some of the design decisions that we have made, and continue to make, concerning our prototype for a high assurance system that evaluates downgrading decisions based upon the amount of private information that may be deduced through inference. Our software system, the Rational Downgrader, is composed of a knowledge-based decision maker to determine the rules that may be inferred, a GUARD to measure the amount of leaked information, and a parsimonious downgrader to modify the initial downgrading decisions. At present, we have restricted the Rational Downgrader to relational databases. Of course, the underlying theories apply to all forms of data. In this paper, we concentrate on design decisions made with the aim of achieving high assurance with respect to an optimality condition.