A Privacy Protection Method for Medical Health Data

Existing works on privacy protection for medical data assume that the values of sensitive attributes are different and ignore the fact that in many real-world hospitals, most patient data are stored and classified by symptoms or departments, which is convenient for the Centers for Disease Control and research institutes to prevent diseases or make decisions. In this paper, we demonstrate that the typical medical scenario can cause serious damage to the privacy of patients’ identities. Motivated attackers can utilize the consistency of sensitive attribute values to perform re-identification attacks, namely, existence attacks. We design an anonymization method based on k-means for protecting medical data from existence attacks. We perform experiments on a real-world dataset to demonstrate that our method can not only resist existence attacks effectively but also lead to less information loss.