Secrecy: Secure collaborative analytics on secret-shared data (BU SYSTEMS SEMINAR)

  • Starts: 12:00 pm on Friday, December 3, 2021
  • Ends: 1:00 pm on Friday, December 3, 2021
In this talk I will present Secrecy, a new relational framework for secure collaborative analytics in untrusted clouds. Secrecy targets challenging use cases where multiple data owners are willing to allow certain SQL queries on their collective private data but do not have private resources to participate in the computation, thus, they need to securely outsource the data analysis to untrusted third parties. To ensure no information leakage and provide provable security guarantees, Secrecy employs cryptographically secure Multi-Party Computation (MPC). MPC achieves strong security via distributing random “shares” of the data across non-colluding entities (e.g., competing cloud providers) and guarantees that no single entity is able to reconstruct the original data. Secrecy’s core novelty is a set of logical and physical optimizations that dramatically reduce query execution costs in the outsourced setting, while retaining the full security guarantees of the underlying MPC protocol. We evaluate Secrecy using real and synthetic queries from several application areas. Our experiments demonstrate that the optimizations we propose can improve query performance by orders of magnitude, enabling Secrecy to outperform state-of-the-art frameworks and scale to much larger datasets than those reported in prior works.

Back to Calendar