
“Program testing can be used to show the presence of bugs, but never to show their absence.”
Edsger W. Dijkstra, Notes on Structured Programming, 1970
Zero-knowledge systems and blockchain cryptographic protocols increasingly operate as production financial and computational infrastructure. Correctness failures can have significant security and financial consequences.
The team applies hard formal methods, protocol modeling, and interactive theorem proving to provide strong correctness assurances for production cryptographic systems. The team works closely with the Ethereum Foundation Protocol Cluster to secure the cryptographic foundations of the L1 proving transition.
The work spans across zk proof systems, zkVMs, protocol verification, execution semantics, and verification infrastructure for production blockchain systems.
Formal verification uses machine-checked proofs to demonstrate that systems behave according to formally specified correctness properties.
Subtle implementation errors, invalid constraints, verifier inconsistencies, or protocol-level failures may not be discoverable through conventional testing or security review alone.
Formal verification complements audits, testing, and security engineering by providing stronger assurances for systems where correctness requirements are unusually high.
Zero-knowledge systems
Cryptographic protocols
Blockchain execution environments
Distributed systems
The team works in established proof tooling and builds its own where the ecosystem has none.
Lean
Halo2
Plonky3
Circom
EVM / Yul
CLAP
CertiPlonk
Halva
Surveyor
ArkLib
Bluebell
The work spans zero-knowledge systems, cryptographic infrastructure, blockchain protocols, and execution environments.
Formal verification of zk proof systems, verifiers, and their implementations, including theFRI formalization in ArkLib and ongoing work toward STIR/WHIR.
Verification of circuit correctness and proving-system behavior across Halo2 and Plonky3, including CLAP, a Lean-based compiler carrying correctness assurances from source specifications to generated constraints.
Mathematical modeling and verification of blockchain protocols and distributed systems.
Reusable verification infrastructure, formal semantics, and theorem-proving tooling for Ethereum and zk systems.
Formal reasoning about smart contract execution behavior for systems requiring stronger correctness assurances. Built on the team's executable EVM and Yul semantics in Lean, the same reference model used for execution-level verification.
A significant portion of the work focuses on reusable verification infrastructure rather than one-off engagements. The team is building formal methods infrastructure for Ethereum and the broader ecosystem.
Every engagement is shaped around the specific properties, assumptions, and threat models of the target system. There is no template. The team works directly with each client to identify the correctness and security properties that matter most, then scopes a verification effort to those failure modes.
The methodology favors the strongest available techniques over the most convenient ones. Specifications are derived from deep technical discussions with teams rather than informal documentation.
The objective is a machine-checked proof, not confidence accumulated from test cases.
Representative verification work across zk systems, execution environments, protocol verification, and formal methods infrastructure.
First formal proof of honesty for a production zk verifier implementation using EasyCrypt.
Verification work across zkVM systems including SP1, Pico, and OpenVM.
Formal semantics of the EVM and Yul in Lean for the Cancun hard fork, passing 99.99% of official Cancun execution tests.
First executable model of the FRI protocol in Lean, with formalization of STIR and WHIR in progress. Building formally verified implementations of these key cryptographic protocols to secure Ethereum during the L1 proving transition.
Framework for extracting and formally verifying Plonky3 zk constraints in Lean without modifying developer workflows.
Lean-based infrastructure for verifying Halo2 circuits used in production zk systems.
Lean-based zk circuit compiler carrying correctness assurances from source-level specificationsthrough generated constraint systems.
Formal verification of key security and correctness properties for the INTMAX2 protocol.
Formal verification is typically applied when systems require stronger correctness assurances than testing or audits alone can provide.
zk proof systems & zkVMs
Blockchain protocols
Cryptographic infrastructure
Execution environments
High-value smart contract systems
Systems secure significant value
Protocol correctness is foundational
Cryptographic assumptions are critical
Infrastructure dependencies expand
Failures are difficult to detect through testing alone
Formal verification research and infrastructure development across Ethereum and zk ecosystems.
Reusable zk verification infrastructure
zkVM verification systems
zk proof-system verification
Post-quantum cryptography
Probabilistic program logics
Contributions also include formalization efforts related to post-quantum cryptography and emerging cryptographic standards.
How formal verification differs from an audit, what it proves, and what an engagement produces.
Audits and testing check for issues someone thought to look for. They can show that bugs exist but never prove that none remain. Formal verification uses machine-checked proofs to show a system behaves according to formally specified correctness properties across every possible execution, which is a stronger assurance for the properties that get verified.
No, formal verification does not replace an audit. It establishes that a system meets the correctness properties it was specified against. Problems outside those specifications, such as access-control misconfigurations or economic design flaws, still need an audit and broader security review. The two are complementary.
A proof establishes that a system matches the correctness properties it was specified against, across all executions those specifications cover, checked by machine rather than by example. The strength of the result depends on the properties specified and the assumptions they rest on, which is why scoping and specification development are central to every engagement.
Each engagement starts with detailed analysis of the system and the properties it needs to hold, followed by threat modeling, specification development, verification planning, formal modeling, and machine-checked proof development. Specifications come from direct technical discussion with the team building the system, not from informal documentation.
Formal specifications, correctness proofs, protocol models, a verification report, remediation guidance, and, in certain cases, reusable tooling.
Work spans zk proof systems and verifiers, zk circuits, zkVMs, cryptographic protocols, blockchain protocols and distributed systems, execution environments, and high-value smart contracts. We are working with several ecosystems, such as Ethereum, Aptos, and ZKsync, and clients such as Succinct, Axiom, Brevis, RISC Zero, and INTMAX.
The Formal Verification team primarily uses Lean, with verification work across Halo2, Plonky3, Circom, and EVM/Yul, supported by in-house infrastructure including CLAP, CertiPlonk, Halva, Surveyor, and ArkLib.
The team's work centers on Ethereum and the EVM, Starknet, and the proof systems and zkVMs used across the ecosystem. Coverage is defined by the proving system and execution environment rather than a fixed chain list, spanning EVM and Yul execution, Halo2 and Plonky3 circuits, zkVMs such as SP1, OpenVM, and RISC Zero, and protocols like ZKsync and INTMAX.