Collaborative Research: SaTC: CORE: Medium: Quicksilver: A Write-oriented, Private, Outsourced Database Management System

Duke University, Durham NC

Investigators

Abstract

Businesses, non-profits, and other organizations are increasingly outsourcing their database needs to third-party cloud providers. The cloud offers many competitive advantages over on-premises deployments including lower costs, high availability, unprecedented scalability, and ease of deployment and maintenance. At the same time, the rise of a hybrid transaction and analytics processing (HTAP) systems presents an additional promise to address the need for real-time data-driven business intelligence by supporting both transactional as well as analytical functions in a single platform. Privacy, legal, and political constraints, however, require organizations to minimize and formally quantify the information leaked outside their organizational boundaries. While there has been much work on building query answering systems on untrusted hardware using encrypted databases and multiparty computation, almost none of these efforts handle database updates, support transactions with interleaving operations, or offer recovery from failures while ensuring strong privacy guarantees. This work will exploit synergies among four areas of research -- databases, oblivious RAMs, secure multiparty computation, and differential privacy. Quicksilver will make it possible to query sensitive data while outsourcing database operations to untrusted cloud providers and this work will have immediate implications for healthcare, federal statistical agencies such as the US Census Bureau, finance, and education. The investigators will integrate this research into a comprehensive education, dissemination and outreach plan that will result in (a) new graduate and undergraduate and graduate courses with open-source materials, (b) the mentoring of graduate students -- especially women and URMs -- on techniques for thriving during their studies, (c) open-source lesson plans for high school teachers on data science, and (d) courses for employees of federal agencies on these topics. In this project, the investigators will design, implement, and evaluate principled techniques for data processing on untrusted third-party platforms. The investigators will first design and implement efficient protocols for updating databases using techniques in cryptography, differential privacy, and systems. They will then create and evaluate algorithms for privacy-preserving concurrency control so that Quicksilver’s transactions will offer interleaving transactions with atomicity, consistency, isolation, and durability (ACID), the gold standard of database systems. In addition, this work will result in protocols for transaction recovery under secure computation when individual cloud nodes fail mid-transaction. This research will reveal novel techniques in oblivious query processing, secure multiparty computation, and differentially-private algorithms. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.