FABLE: Batched Evaluation on Confidential Lookup Tables in 2PC
summary
FABLE is a concrete system that implements confidential lookup table evaluation by combining batch PIR, symmetric encryption, and batching to get sublinear client work and sublinear communication. The implementation encodes the server table into buckets, offloads cuckoo hashing to the client via OPRF, caches duplicate-removal work to save 84.1% communication versus prior baselines, and adds database masking plus noise flooding for confidentiality. Evaluation shows up to 28–101× LAN and 50–393× WAN speedups, >400× on embedding lookup, and ~15× on a join-style baseline.
AI-written summary · about
chapters (11)
- 0:00FABLE protocol intro
- 2:44Lookup table use cases
- 5:13Problem and design goals
- 6:17Naive PIR construction
- 8:54Batching for performance
- 10:22Handling batch duplicates
- 11:27Client-side cuckoo hashing
- 12:11Lookup table confidentiality
- 12:56Performance vs prior works
- 13:16Real-world benchmarking
- 13:59Conclusion and future
related talks
23mBatchCrypt: Efficient Homomorphic Encryption for Cross-Silo Federated Learning
38mDemystifying Privacy Preserving Computing
18mSECRECY: Secure collaborative analytics in untrusted clouds
01h 08mSecure Computation 1
27mMetadata-private data transfer
25mFaster Malicious 2-Party Secure Computation with Online/Offline Dual Execution
01h 14mPrivate Information Retrieval and Oblivious RAM
44mIP Protection and Privacy in LLM: Leveraging Fully Homomorphic Encryption