Abstract

Data sharing to support research or other purposes can be highly beneficial but may also entail serious privacy risks when the shared data refers to individuals or can lead to the disclosure of sensitive knowledge patterns, such as trade secrets, when mined. To overcome these threats, the research areas of privacy-preserving data publishing and privacy-preserving data mining were brought into existence. As is well-known, the offering of privacy comes at a cost to data utility. Consequently, achieving a good balance between privacy and data utility is fundamental when designing privacy-preserving algorithms.

In this talk, I will provide an overview of my recent research work in data anonymization and knowledge hiding, with emphasis on anonymization methods for medical data sharing. These methods can increase data availability and utility, which is important for supporting research on personalized medicine. This is because they provide privacy guarantees, as well as they allow biomedical tasks, such as genome-wide association studies and clinical case analysis, to be performed accurately. To demonstrate the quality of the anonymization approaches, I will present a case study using data from the EMR system of the Vanderbilt University Medical Center (VUMC), a state-of-the-art system that stores information about 2 Million patients over 15 years.