Health Data Management and Analytics with Privacy and Confidentiality

摘要

Managing and analyzing large-scale clinical and public health data while protecting privacy of human subjects has been a key challenge in biomedical research. Traditional de-identification approaches are subject to various re-identification and disclosure risks and do not provide sufficient privacy protection for patients. This talk gives an overview of our work on privacy preserving health data sharing and analytics along two dimensions: (1) data encryption techniques that support secure computation and query processing on the encrypted data without disclosing the raw data, (2) differential privacy techniques that ensure the computation and query results do not disclose patient information. Focusing on the second dimension, a set of differentia] privacy techniques are presented that handle different types of data including relational, sequential, and time series data. Case studies using real health datasets are presented to demonstrate the feasibility of the solutions while outlining their limitations and open challenges.