Designing Tunable and Scalable Oblivious Databases

Abstract: Today, over 94% of enterprises - including 83% of healthcare organizations - rely on cloud computing to outsource their data storage and computation needs due to the many benefits of the cloud. However, outsourcing plaintext data to third-party providers can expose sensitive information to potentially untrustworthy entities. While encryption helps protect data at rest and in transit, it falls short against attacks that exploit access patterns on encrypted data to infer sensitive information.

Oblivious databases address this vulnerability by hiding access patterns during data operations. Yet, existing solutions often suffer from significant limitations: high storage and bandwidth overheads, lack of scalability, and restricted query capabilities. In this talk, I will introduce Waffle, a tunable-oblivious algorithm that achieves constant storage and bandwidth overheads. Building on Waffle, I will present OasisDB, a scalable and oblivious database framework that supports rich relational queries while maintaining strong privacy guarantees.

Bio: Sujaya Maiyya is an Assistant Professor at the University of Waterloo. Her research interests lie in the intersection of data management, distributed systems, and data privacy. Dr. Maiyya has authored papers at top venues, such as VLDB, SIGMOD, and USENIX Security. During her PhD, Dr. Maiyya received the IBM PhD Fellowship for her contributions in large-scale data management. She is also a core member of Diversity & Inclusion in Databases initiative, striving for DEI efforts in the database community.