Abstract

The problem of data privacy is to verify that confidential information stored in an information system is not provided to unauthorized users and, therefore, personal and other sensitive data remain private. One way to guarantee this is to distort a knowledge base such that it does not reveal sensitive information. In the present paper we will give a universal definition of the problem of knowledge base distortion. It is universal in the sense that is independent of any particular knowledge representation formalism. We will then present a basic and general algorithm for knowledge base distortion to guarantee data privacy. This algorithm provides us with upper bounds for the complexity of the distortion problem. Moreover, we examine heuristics to improve its average performance