Binius: a Hardware-Optimized SNARK
Jim Posen
Abstract:
Verifiable computing with zk-SNARKs is taking off as a solution to the scalability and privacy limitations that today's blockchains suffer from. Binius is a novel succinct argument (SNARK) constructed using field extension towers of binary fields. Its design is uniquely suitable for the critical use case of verifying classic and widely deployed symmetric cryptography primitives like SHA-256 and Keccak-256. Efficient arithmetization of traditional hash functions unlocks the recursion composition of Binius proofs without relying on less battle-tested and slower arithmetization-optimized hash functions like Poseidon. Binius builds on recent works using multilinear polynomial IOPs to construct SNARKs, such as HyperPlonk, Brakedown, and Lasso. The core innovation is a new polynomial commitment scheme (PCS) where committing to field elements as small as single bits is no more expensive than committing to the same amount of information represented with larger field elements. The PCS achieves this using a technique inspired by concatenated error-correcting codes. Beyond the arithmetization benefits of the Binius approach, the finite field operations lend themselves to highly efficient implementation in native digital logic.
This talk covers the fundamental ideas behind Binius, published in the paper "Succinct Arguments over Towers of Binary Fields", as well as a performance comparison of Ulvetanna's software and hardware implementations of primitive SNARK operations in Binius with the analogous operations in prime field-based SNARKs. This talk is based on work by Ulvetanna, a company building FPGA-accelerated compute clusters for verifiable computing.
Bio:
Jim Posen is the co-founder and Cryptography Engineering Lead at Ulvetanna, a cryptographic acceleration company focused on verifiable computing. Previously, Jim was an early engineer at Coinbase, where he was the Tech Lead on the Payments team responsible for bank and blockchain integrations. Jim also established the Protocol team at Coinbase, and contributed to open-source projects serving Bitcoin, Ethereum, and the Lightning Network. Jim has a BSE in Electrical and Computer Engineering from Duke University and an MS in Computer Science from Stanford University.