Weave - Efficient and Expressive Oblivious Analytics at Scale

Learn about Weave, an innovative analytics platform designed to perform secure computations on sensitive data at scale in this 15-minute conference presentation from OSDI '25. Discover how researchers from Yale University address the critical challenge of access pattern attacks in cloud-based distributed analytics, even when using trusted hardware enclaves and encryption. Explore Weave's novel approach that combines noise injection with hardware memory isolation through enclave page caches to achieve constant-factor network and compute overheads, significantly improving upon previous solutions that suffered from logarithmic scaling penalties. Understand the platform's optimizations and extensions that leverage dataset and workload-specific properties to maintain both performance and functionality without compromising security. Examine evaluation results demonstrating 4-10× improvements in end-to-end execution time across various analytics jobs on large real-world datasets compared to existing state-of-the-art solutions, all while providing robust obliviousness guarantees against access pattern leakage attacks.