Ulvetanna's co-founders, Jim Posen and Radi Cojbasic, discuss the origin story of Ulvetanna and their work on the ZK hardware/software intersection. They delve into Binius, their new proving system optimized for hardware, which draws on recent breakthroughs in SNARKs. Topics include smaller fields, new lookup arguments, multilinear provers, and recursive composition in SNARKs.
Read more
AI Summary
Highlights
AI Chapters
Episode notes
auto_awesome
Podcast summary created with Snipd AI
Quick takeaways
Ulvitana focuses on accelerating zero-knowledge proofs through hardware-software code design and emphasizes the need for dedicated custom compute for efficient adoption.
Binius is a new proving system that uses binary fields, multi-linear protocols, lookup arguments, and recursive composition to overcome limitations in previous proof systems and adds flexibility to constraint systems.
Binius leverages breakthroughs in cryptography to create a Snark over binary fields, eliminating embedding overhead and enabling efficient arithmetic, memory usage, and wider adoption of verifiable computing.
Deep dives
Ulvitana's Work on ZK and Hardware Intersection
Ulvitana is a company focused on accelerating zero-knowledge proofs through hardware-software code design. They aim to decompose proofs into CPU-friendly and hardware-friendly components. By using FPGAs for their Gen1 server, they prioritize reconfigurability, allowing them to adapt quickly to the fast-paced innovations in zero-knowledge space. They also emphasize the need for dedicated custom compute for the full adoption and efficiency of zero-knowledge proofs.
Binius: A New Proving System Based on Binary Fields
Binius is a new proving system developed by Ulvitana. It uses binary fields, which are finite fields of characteristic two, and offers advantages in terms of performance, security, and compatibility with standard hash functions. By combining the use of binary fields with multi-linear protocols, lookup arguments, and recursive composition, Binius overcomes limitations seen in previous proof systems. The introduction of smaller fields and the ability to choose between different field sizes also adds flexibility to the constraint systems. Ulvitana has released a software library for Binius, which marks an important step towards implementing and deploying this innovative proving system.
Challenges and Advantages in Hardware-based Approaches
Accelerating zero-knowledge proofs using hardware presents challenges such as data movement between the CPU and accelerator, bandwidth limitations, and the need to adapt to evolving cryptographic algorithms. Ulvitana's research and development efforts focus on addressing these challenges and improving hardware efficiency. By exploring alternatives like binary fields and leveraging advancements in sum-check protocols, multi-linear provers, and field extension techniques, Ulvitana aims to provide efficient and verifiable computing solutions. Additionally, adopting more widely accepted and performant hash functions like Grostl offers benefits in both security and performance.
Binias: A Snark over Binary Fields
Binias is a proof system that utilizes breakthroughs in cryptography to create a Snark over binary fields. The key contribution of Binias is its ability to eliminate embedding overhead, which has been a challenge in previous protocols like Starks. By using binary fields and their tower construction, Binias allows for efficient F2 operations without the need for padding with zeros. This not only provides computational efficiency but also reduces system memory usage and enables more efficient arithmetic. Binias is considered a Snark and has the potential to be a platform for ZKVM implementations and contribute to the wider adoption of verifiable computing.
Hardware Efficiency and the Future of Proof Systems
Binias leverages the efficiency of binary field operations in hardware, which have been extensively studied in cryptographic applications like the AES encryption scheme. By incorporating hardware optimization early in the R&D phase, Binias ensures computational efficiency and higher clock frequencies, resulting in better overall performance. This hardware and software code design approach is crucial in achieving cost-efficient and high-performance proof systems. The use of binary field data types, which align with common computing practices based on power-of-two data types, further enhances performance benefits. The combination of algorithm, hardware, and software design allows for the development of not just efficient Snarks, but also lays the foundation for broader applications in verifiable computing.
In this week's episode, Anna and cohost Brendan Farmer catch up with Jim Posen and Radi Cojbasic from Ulvetanna. They cover the origin story of Ulvetanna and their work on the ZK hardware/software intersection before moving on to discuss Binius, a new proving system they developed which is optimised for hardware. Binius is built on towers of binary fields and draws on recent breakthroughs on SNARKs. This work continues the trend towards the use of smaller fields and was inspired by the development of new lookup arguments, work done on multilinear provers and sum-check as well as the use of recursive composition in SNARKs.
