An Open Source Tool for Game Theoretic Health Data De-Identification.

Area of Science:

• Health Informatics
• Data Privacy
• Computational Social Science

Background:

• Biomedical data is rapidly expanding, necessitating secure data sharing for research.
• Current de-identification methods often over-corrupt data due to worst-case adversary assumptions.
• Game theory offers a promising framework for understanding data publication incentives but lacks practical application.

Purpose of the Study:

• To develop and implement a practical game theoretic data publication strategy.
• To integrate this strategy into the open-source ARX software for broader accessibility.
• To evaluate the strategy's performance regarding data utility, privacy, and efficiency.

Main Methods:

• Developed a novel game theoretic model for data publication.
• Integrated the model into the ARX open-source software.
• Evaluated the implementation using over 30,000 U.S. Census Bureau demographic records.

Main Results:

• The implemented game theoretic strategy is scalable for large datasets.
• Demonstrated a practical balance between data transformation, utility, and privacy.
• Showcased the integration's compatibility with diverse privacy risk and quality metrics.

Conclusions:

• Game theoretic approaches can be practically applied to enhance secure biomedical data sharing.
• The ARX software integration provides a valuable tool for researchers and data publishers.
• This work advances the field by bridging theoretical game theory with real-world data privacy challenges.