By now, everyone into zk knows about lookup tables (thanks to Plookup, Lasso, etc.). But did you know that lookup tables (LUTs) also exist in MPC?

Why explore MPC lookups? They help balance a common trade-off in MPC: secret sharing-based protocols have minimal total communication but require many rounds, while Yao-Garbled Circuits (GC) have a constant number of rounds but higher communication overhead.

The most compelling application of LUTs is private ML inference (more on this in another post). However, there's an underexplored area: MPC LUTs for collaborative SNARKs.

To support proof systems that leverage LUTs, we need to find efficient ways to compute zk LUTs in an MPC-fied manner. We are actively researching this area and would love to speak with anyone interested in this topic.

Building at TACEO
Believing in programmable cryptography.

We use garbled circuits with oblivious transfer for MPC-TLS. Networking is the main bottleneck.

Working on updates now that switch to use vector oblivious linear evaluation.

This lets us construct an efficient 1 of N oblivious transfer. So we can use lookup tables.

Brings down overhead by a factor of 100 😉

Working on building zkTLS infra/products. A ruthless empiricist and meme connoisseur.

http://opacity.network

Nice. 100 is huge.

Pls share once you can.