Privacy-Preserving Biomedical Database Queries with Optimal Privacy-Utility Trade-Offs

Area of Science:

• Biomedical Informatics
• Data Privacy
• Computational Biology

Background:

• Interactive query-answering systems are crucial for sharing aggregate insights from biomedical databases.
• Existing systems risk leaking sensitive individual-level data despite aiming for privacy.
• Differential privacy offers a framework for formal privacy guarantees.

Purpose of the Study:

• To develop query-answering mechanisms that maximize utility (accuracy) while ensuring formal privacy guarantees.
• To extend theoretical results on differential privacy to more general utility functions relevant to biomedical research.
• To demonstrate improved accuracy over existing methods for various biomedical data analysis tasks.

Main Methods:

• Application of recent advances in differential privacy to design query-answering mechanisms.
• Theoretical analysis to extend the proof of optimality for differential privacy mechanisms.
• Empirical evaluation on use cases including cohort discovery, variant lookup, and association testing.

Main Results:

• Demonstrated accuracy improvements over existing approaches for biomedical data analysis.
• Extended the theoretical optimality of differential privacy mechanisms to general utility functions.
• Achieved provable privacy guarantees while maximizing system utility.

Conclusions:

• The developed mechanisms offer an optimal balance between privacy and utility in interactive biomedical databases.
• This work provides a pathway for secure and efficient data sharing in biomedical research.
• The findings are applicable to critical workflows like cohort discovery and association testing.