Enhancing utility and privacy-safety via semi-homogenous generalization

Abstract

The existing solutions to privacy preserving publication can be classified into the homogenous and non-homogenous generalization. The generalization of data increases the uncertainty of attribute values, and leads to the loss of information to some extent. The non-homogenous algorithm which is based on ring generalization, can reduce the information loss, and in the meanwhile, offering strong privacy preservation. This paper studies the cardinality of the assignments based on the ring generalization, and proved that its cardinality is α n (α > 1). In addition, we propose a semi-homogenous algorithm which can meet the requirement of preserving anonymity of sensitive attributes in data sharing, and reduce greatly the amount of information loss resulting from data generalization for implementing data anonymization. © 2012 Springer-Verlag.