Abstract

k-Anonymity and l-Diversity have laid the fundamental techniques for preserving privacy in microdata, and many research works have been inspired by them, proposing better and stronger levels of privacy. A common technique for achieving higher privacy in microdata tables is to diversify the records in such a way that sensitive information stored in the data is less likely to be disclosed. While most of the approaches succeed in protecting the original sensitive information to a high degree, issues arise when sensitive values are generalised along a hierarchical taxonomy, causing an increase in probability of privacy disclosure already after the first level of generalisation. This paper introduces n-Dependency, a novel technique that considers the hierarchical nature of sensitive information and their generalisations when diversifying the microdata. We propose a formal model and algorithms, and verify our technique by conducting extensive experiments