Hide Your Distance: Privacy Risks and Protection in Spatial Accessibility Analysis.

Area of Science:

• Public Health
• Geospatial Analysis
• Data Privacy

Background:

• Assessing spatial accessibility to healthcare is crucial for public health, especially during disease outbreaks.
• Sharing individual travel data for accessibility analysis raises significant privacy concerns, including potential disclosure of personal information.
• Existing methods lack robust privacy protections for sensitive spatial health data.

Purpose of the Study:

• To investigate privacy risks, specifically membership inference and address inference, associated with sharing spatial accessibility data.
• To develop and evaluate novel, provably private algorithms for spatial accessibility analysis.
• To ensure the utility of spatial accessibility data while mitigating privacy disclosures.

Main Methods:

• Investigated information leakage through membership inference and address inference in spatial accessibility computations.
• Developed two novel algorithms with provable privacy guarantees.
• Evaluated algorithms using real-world population and healthcare facility data from Mecklenburg County, NC, and Nashville, TN.

Main Results:

• The proposed algorithms effectively reduce the risks of membership and address disclosure compared to state-of-the-art privacy practices.
• Demonstrated that privacy-preserving spatial accessibility analysis can be achieved without significant loss of data utility.
• Empirical validation on real datasets confirmed the effectiveness of the developed privacy-preserving methods.

Conclusions:

• Sharing spatial accessibility data poses inherent privacy risks that can be mitigated through advanced algorithmic approaches.
• The developed provably private algorithms offer a viable solution for privacy-preserving spatial accessibility analysis in public health.
• These methods enable safer data sharing for informed public health policy and intervention design.