Treebeard: A Scalable and Fault Tolerant ORAM Datastore
Amin Setayesh
34th USENIX Security Symposium (USENIX Security '25) · Day 3 · Privacy 4: Privacy-Preserving Computation
In the realm of data privacy, encryption has long been considered the gold standard for protecting sensitive information. However, this talk, "Treebeard: A Scalable and Fault Tolerant ORAM Datastore," presented by Amin Setayesh at USENIX Security, challenges this conventional wisdom by highlighting a critical vulnerability often overlooked: **access pattern leakage**. Even when data is fully encrypted and outsourced to a cloud database, the patterns of how users interact with that data can inadvertently reveal significant insights to an adversary. This talk introduces Treebeard, a novel **Oblivious RAM (ORAM)** system designed to address these fundamental privacy gaps, offering a robust solution that is not only secure but also scalable, fault-tolerant, and performant.
AI review
Treebeard solves a real and underappreciated problem — access pattern leakage in outsourced storage — and the architectural contributions (decoupled Router/Stash/ORAM layers, multipath reads/eviction, Raft-backed fault tolerance) are technically legitimate and non-trivial. This is solid systems security research that advances the ORAM practicality problem meaningfully, but it's firmly a conference paper presentation rather than a field-defining talk, and the novelty ceiling is limited by how niche practical ORAM deployment remains.