SECRECY: Secure collaborative analytics in untrusted clouds
summary
SECRECY implements cryptographic MPC with arithmetic and Boolean secret sharing, a co‑designed query engine and physical optimizations that batch messages to cut synchronization rounds to O(k)+1. It models cost as operational, synchronization and composition components to pick plans that lower communication, space and work. Multi‑cloud experiments report over 1000x improvement versus prior work, handle up to 2 million rows, and run the heaviest query in about one hour while yielding up to 38x lower latency on some queries.
AI-written summary · about
chapters (9)
- 0:00SECRECY framework intro
- 2:33Secure collaborative analytics challenges
- 4:59SECRECY vs prior approaches
- 6:25SECRECY’s secure SQL execution
- 9:18Core optimizations in SECRECY
- 11:47Cost model for query plans
- 14:16SECRECY’s vectorized engine
- 15:42Performance and use cases
- 17:10Q&A
related talks
01h 31mIntroduction to Secure Multi-Party Computation
01h 08mSecure Computation 1
45mConfidential Computing: Protecting Customer Data in the Cloud
12mLessons from years of building privacy-preserving technologies
16mUnlocking Cross-Organizational Insights: Practical MPC for Cloud-Based Data Analytics
10mKeynote: A Secure Collaborative Learning Platform
25mPractical Accountability of Secret Processes
40mConfidential Computing and Privacy-Enhancing Technologies - The Landscape