Modeling Background Knowledge for Privacy Preserving Medical Data Publishing

Currently, many privacy preserving schemes for medical data publishing faces the potential threats of background knowledge based attacks, however, the principle of the attacks has not been fully studied. How background knowledge impacts on the privacy disclosure for various of privacy preserving schemes is still a new research topic to be solved. In this paper, we study the background knowledge based attacks and propose a model to quantify background knowledge which is used to infer patient privacy. Besides, we simulate three popular anonymity algorithms (K-anonymity, L-diversity, t-closeness) on sample datasets and testify our background knowledge attack model. We believe that our research could help us to understanding the impact of background knowledge on privacy inference of published medical data.