H2O2RAM: A High-Performance Hierarchical Doubly Oblivious RAM
summary
H2O2RAM is a hierarchical doubly oblivious RAM construction that replaces scattered tree layouts with contiguous level tables, implements three optimized oblivious hash components (PRF-based bucket hashing, a k-candidate stashless cuckoo variant with at most 6 oblivious comparisons, and a two-tier sampled hash), and parallelizes rebuilds. Evaluation shows an implemented system and open source code, reporting up to ~10^3× speedup and 5–44× memory savings versus prior O2RAMs on realistic workloads.
AI-written summary · about
chapters (11)
- 0:00H2O2RAM intro and ORAM challenges
- 1:59Confidential computing with ORAM
- 3:05RAM locality and performance
- 4:11H2O2RAM design and hash schemes
- 6:42Optimal bucket count
- 7:45Reducing stash size
- 9:37Secondary hash table and input distribution
- 11:07Hash schemes for input sizes
- 12:10Performance vs state-of-the-art ORAM
- 13:13Experimental validation and speedup
- 14:18Conclusion and resources
related talks
30mUsing Data-Oblivious Algorithms for Private Cloud Storage Access
33mYes, There is an Oblivious RAM Lower Bound!
23mOblivP2P: An Oblivious Peer-to-Peer Content Sharing System
01h 14mSafe user-level sharing of memory-mapped resources
15mORC: Increasing Cloud Memory Density via Object Reuse with Capabilities
15mOsprey: Transparent and Efficient Virtual Memory for Secure Computation
40mRowhammer Attacks: A Walkthrough Guide
16mPatronus: High-Performance and Protective Remote Memory