KOSUNI CRYPTO
Korea's Crypto Pulse, in English
$BTC $ETH $XRP $SOL $DOGE
News
Neutral

Five Verification Paths to Reduce Ethereum Client Trust-Computing Base

Published September 25, 2026 9:28 PM · 10 views $ETH
Five Verification Paths to Reduce Ethereum Client Trust-Computing Base

Ethereum Foundation member George Kadianakis and Ethereum Foundation member Kev Wedderburn published a post on the Ethereum Research Forum on May 25 detailing five design and implementation paths for formal verification aimed at reducing the trust computing base (TCB) of Ethereum clients. The TCB refers to components, specifications, tools, and assumptions that are trusted without verification.

Formal verification cannot eliminate the TCB entirely but can reduce the scope requiring manual trust. The research team proposed dividing the abstracted client into modules with separate roles and guarantees, connecting module guarantees via interfaces to verify overall security properties.

The key distinction is between 'pure modules' and 'non-pure modules.' Areas like cryptography, SSZ, and fork choice rules—having minimal side effects and clear mathematical structures—are classified as pure modules suitable for formal verification. Network-related areas with high input/output complexity and considerations of message order, disconnections, and delays are considered non-pure modules.

The researchers suggested designing non-pure modules to not be trusted from the outset. For example, instead of directly trusting signatures delivered by network modules, they should be reprocessed through a pure signature verification module to handle bugs or malicious inputs similarly. Long-term, they proposed moving verification boundaries closer to the network—verifying parsers, gossip rules, and synchronization logic before modeling the entire network to prevent malformed messages from halting systems or consuming excessive resources.

The post also identified connecting specifications with actual executable code as a separate challenge. The team distinguished between writing formal specifications in Lean4 and proving properties, and verifying that implementation adheres to those specifications. Users prioritize their computer's running code safety over Lean4 theorems, making both steps necessary.

Five implementation paths were proposed: 1) Automatically converting Rust-written code to Lean4; 2) Extracting Lean4 modules into C code; 3) Writing most of the client in Lean4 and connecting high-side-effect modules via other languages; 4) Directly writing core modules in RISC-V assembly; and 5) Using verified compilers.

Converting Rust to Lean4 leaves the translator and final binary compiler within TCB. Extracting Lean4 to C or writing most of the client in Lean4 includes extractors, C compilers, and external function interfaces as trusted components. Writing core modules in RISC-V assembly excludes general compilers from TCB but introduces new trust targets like formal models of RISC-V instruction sets and tools translating code for other processors. Verified compilers allow excluding extractors and C compilers from TCB.

The researchers emphasized that not all modules need the same approach—mathematically strong modules can be verified in Lean4 while complex data structures or large-scale modules may use translators or verified compilers, allowing hybrid approaches. This post does not announce that Ethereum clients are already fully formally verified but presents a design direction for how verification might be scaled across different units of implementation.

Korean Source

This article is an English localization of a Korean-language crypto news report. Original headline: 이더리움 클라이언트 신뢰 기반 줄일 검증 경로 5가지