Privacy-Preserving Biomedical Data Sharing

点击翻译按钮获取中文摘要

The adoption of Electronic Health Record (EHR) systems has accelerated the collection and sharing of health data. EHR databases are often integrated with biorepositories and shared to facilitate genome-wide association studies (GWAS). While such integration and sharing of biomedical data have many social benefits (e.g., personalized medicine), it also raises concerns about individual privacy. A recent survey on privacy attacks demonstrates various techniques for breaching biomedical data privacy, and these techniques question the efficacy of the protection mechanisms that are currently in practice. The aim of this research is to understand, evaluate and contribute to the problem of sharing biomedical data while protecting individual privacy. The research activities are organized along the three related tasks. The first task is to more fully understand the privacy threats and challenges of sharing biomedical data, and to propose appropriate privacy models. The challenge is to thwart privacy attacks caused by inferences between different data types, and to quantify the risk of data sharing for a given adversary model. The second task is to develop data-sharing mechanisms with strong privacy guarantees. The challenge is to provide a reasonable trade-off between privacy and data utility. The final task is to develop distributed mechanisms to securely integrate data from multiple institutions to facilitate controlled and effective data sharing. The challenge is to propose a scalable algorithm that safeguards the raw data stored across multiple institutions against an untrusted data aggregator. The success of biomedical research heavily depends on the availability of large datasets. At present, there is no specific national guideline for sharing biomedical data. The outcomes from this research will help to bridge this gap by providing clear guidance and technical solutions, and will help increase public trust in biomedical research. Further, this research will advance the current state-of-the-art techniques for privacy-preserving data sharing and broaden our understanding of the privacy threats caused by the interconnections of different types of data.