Utility-Prioritized Differential Privacy for Quantitative Biomedical Data

Abstract

In the context of heavy investments in life sciences, one consequence of rapid biotechnical developments is the generation of very large amounts of biological data that not only support bioinformatics research used in medical and health services by data analysis but also cause genetic privacy breaches for the data originators, limiting biomedical data sharing. However, the original intention of medical informatization is to improve medical efficiency and break the information barriers between medical institutions, which is also limited by genetic privacy risks from growing bioinformatics data. To share biomedical data correlated with genes usefully under the premise of ensuring genetic privacy, with the concept of medical reference interval, we design a query strategy based on a flexible differential privacy algorithm that setting weights for noise added according to the purpose of the inquirer. We generate a series of biomedical data to test the effectiveness of our query system, the test results show that our algorithm ensures data’s utility with acceptable privacy breaches that addresses the problem of biomedical data sharing to a certain extent.