A scheme for inference problems using rough sets and entropy

Abstract

The inference problem is an unauthorized disclosure of sensitive information via indirect accesses. It happens when generic users can infer classified information from the data or relations between data in a dataset available to them. This problem has drawn much attention from researchers in the database community due to its great compromise of data security. Unlike previously proposed approaches, this paper presents a new scheme for handling inference problems, which considers both security and functionality of a dataset. The scheme uses two main tools. One is the application of rough sets to form a minimal set of decision rules from the dataset. The other is the use of entropy, an important concept from information theory, to evaluate the amount of information contained in the dataset. By analyzing the changes of confidence in decision rules and in the amount of information, an optimal solution can be decided. The scheme is explicit and also easy to be implemented. © Springer-Verlag Berlin Heidelberg 2005.