A Federated Query Optimizer for Privacy-Preserving Analytics
Data sharing within DHS agencies and across components is challenging because of both privacy and security concerns. Many agencies remove sensitive but critical information from reports or documents to preserve privacy while others refrain from sharing data that could be useful to homeland security efforts due to the challenges of meeting and maintaining privacy requirements.
This project is developing a novel framework to automatically recommend privacy-preserving methods using a federated query optimizer. A federated approach allows different siloed systems and applications to solve for specific functions independently but within the same set of standards. The optimizer will automate privacy-preserving analytics that will allow the data to be used from different systems while protecting privacy. Using learning-based technology, the system will produce query results that achieve a balance between efficiency and privacy. The system will also differentiate the data shown based on the user’s role and data access permissions. Models can be refined by the users to work with preferred domain applications.
This approach will ease operational efforts on data protection by automating privacy-preserving analytics so that the system automatically protects private information once such information is structured and identified. Based on learning-based mechanisms, the system will develop query results that achieve a balance between efficiency and privacy. The results of this research will allow for more accurate data analysis within and between organizations while meeting privacy requirements.
Research Leadership Team
Principal Investigator: Jia Zou, Arizona State University
Co-Investigator: Chaowei Xiao, Arizona State University
Co-Investigator: Yingzhen Yang, Arizona State University