Privacy Preservation in Publishing Electronic Health Records Based on Perturbation

Abstract

The patients’ health information is often kept as electronic health records (EHRs). To improve the quality and efficiency of the care, EHRs can be shared among different organizations. However, the inappropriate sharing or usage of these healthcare data could threaten people’s privacy. It becomes increasingly important to preserve the privacy of the published EHRs. An attacker is apt to identify an individual from the published EHRs by partial measurement information as background knowledge, with attacks through the record linkage and attribute linkage. To resist the above types of attacks, we propose a privacy preservation with perturbation in the published healthcare data (PPHR). To protect the privacy of sensitive information, we first determine the critical sequences based on which some specific records are easy to be identified. Then, we adopt perturbation on these sequences by adding or deleting some points while ensuring the published data to satisfy l-diversity. A comprehensive set of real-life healthcare data sets are applied to evaluate the performance of our anonymization approach. Simulations show our scheme possesses better privacy while ensuring higher utility.