Equilibrium Infra Bulletin #31: Murky Proof System Waters, State of ZK, Plonky3, and more...
Equilibrium Labs builds the state-of-the-art of decentralized infrastructure. We are a global team of ~30 people who tackle challenges around security, privacy, and scaling.
Infra Bulletin will be on summer break until the 15th of August ☀️
🔍The Murky Proof System Waters (Part 1)
⚡️ Topic Summary
At the heart of every zkVM is the proving scheme that turns a trace of a computation into an efficiently verifiable proof of its execution. The choice of proof system is one of the defining design decisions for a zkVM, as it impacts speed, security, interoperability, and suitability for a variety of applications.
The key properties that we want out of a proving scheme used for zkVMs are:
Performance - Want the proving scheme to be fast to reduce the overhead of running a program in a zkVM vs direct execution on hardware.
On-chain verifiability - Most efforts today focus on making verification cheap and efficient on Ethereum/EVM and for the proof small enough to be stored onchain.
Precompiles - Can significantly boost computationally intensive application domains (such as ML inference) to reach practicality by writing dedicated circuits to avoid the overhead of the CPU abstraction for computations. The tradeoff here is privacy as the verifier can see that a precompile has been used, but this is not relevant for public verifiable computation use cases.
🤔 Our Thoughts
The pace of development for both proving schemes and zkVMs over the last couple of years has been incredible to follow. Performance is improving and the overhead is decreasing, meanwhile, decentralized proving networks such as Gevulot enable easy outsourcing of proof generation.
The downside of speed is that the wider community struggles to keep up with all the new research, particularly novel implementations, which can impact the robustness and our confidence in these schemes. This makes it difficult to decide which scheme to prioritize - both for engineers (robust implementation) and the broader research community (scrutiny and standardization).
In addition, edge cases are still not widely understood and the practice of ZK-auditing is only getting started. There are ways around this, such as taking a multi-prover approach or having other guardrails, but at the end of the day, it’s still novel technology that hasn’t been fully battle-tested yet.
Most efforts today are focused on extending Ethereum and the EVM, but there are some teams that have chosen to optimize for alternative verification environments (Aleo, Mina…). These teams don’t face the same constraints which gives more design freedom but limits interoperability between different systems. At the end of the day, we will need to agree on some standards, but that still seems some time away.
💡 Research, Articles & Other Things of Interest
📚 Nexus 2.0: Jolt, HyperNova, and a New SDK: Integrates Jolt on the frontend, Hypernova recursive proof system on the backend, and CDK framework for producing multiple proofs in parallel and at scale.
📚 Polygon Plonky3 is Production Ready: With fewer fixed properties (“less opinionated”) than Plonky2, developers have a wider variety of choices to build bespoke proving systems for zkVM or zkEVM chains. It’s modular across both finite fields (BabyBear; Mersenne31; Goldilocks) and hash functions (Poseidon; Poseidon2; Rescue; BLAKE3; Keccak-256; Monolith), with plans to add optionality for polynomial commitment-scheme as well.
📚 State of ZK - Q2 2024: The latest report from the ZK Validator team contains research on sum-check protocols, optimizations around small fields, ZK hardware, and benchmarking of validity rollups.
🎧 ZK in the Wild: Building private payments and beyond: Payy is a ZK-based mobile application for private payments (think Venmo, but with private stablecoin transfers). The tech stack is a Polygon-based validium, but it’s all abstracted away to a mobile application with easy onboarding and smooth UX.
🔥 News From Our Partners:
Gevulot announces ZkBoost - Proof Supply Chain Abstraction: The idea of ZkBoost is to be a kind of proxy node that proving services and networks can integrate with - providing a single, standardized API for Layer 2 networks to access all proving options (Unification > Fragmentation). Gevulot is funding early work on ZkBoost, but they aim for it to be neutral, open-source software. Reach out to Nilu if you want to join this effort!
🤌 Personal Recommendations From Our Team
📚 Reading: Tracers in the Dark - Andy Greenberg: Veteran cybersecurity reporter Andy Greenberg covers the stories of criminal empires built and destroyed, including how they tracked down Silk Road, AlphaBay, and more.
🎧 Listening: Through the Fire and Flames - DragonForce (Speed Bag Cover): When you wanted to be a rockstar, but all you got was boxing equipment 🤷♂️ Impressive control of the speed bag in this punch drumming cover nonetheless.
💡 Other: Using Proteins To Assemble Semiconductors: Seems pretty far away from being production-ready, but a cool idea. Time to pivot into biotech? 🧑🔬