Coin Brief ENDE

EIP-8355 adds ML-DSA-65 and ML-DSA-87 test vectors for its post-quantum signature precompiles

A commit merged into the EIPs repository on 7 November added verification vectors and rejection cases for ML-DSA-65 and ML-DSA-87 to EIP-8355, a draft by Danno Ferrin that proposes precompiles for verifying ML-DSA signatures on Ethereum. The update adds about 400 lines to the proposal's test-vector file.

ML-DSA is the lattice-based signature scheme that NIST standardised in FIPS 204, and the EIP's motivation is that Ethereum needs a post-quantum signature check a contract can call. It defines three precompiles, VERIFY_MLDSA44 at address 0x12, VERIFY_MLDSA65 at 0x13 and VERIFY_MLDSA87 at 0x14, for NIST security levels II, III and V. Each takes one concatenated input, public key followed by signature followed by message, with no length prefixes, and returns a 32-byte word set to one for a valid signature and zero otherwise.

The draft sets itself against EIP-8051, which also proposes ML-DSA verification, on two points. EIP-8051 covers only ML-DSA-44; EIP-8355 argues that account keys hold funds for years and should be able to use the stronger levels. And EIP-8051 fixes the message at a 32-byte pre-hash, while EIP-8355 takes a message of any length as the trailing bytes, since FIPS 204 defines verification over arbitrary messages and the key and signature lengths are fixed.

Gas is a base cost plus 6 per 32-byte word of message: 6,500 for ML-DSA-44, 9,000 for ML-DSA-65 and 13,500 for ML-DSA-87, charged in full before any verification. Minimum inputs are 3,732, 5,261 and 7,219 bytes.

EIP-8355 adds ML-DSA-65 and ML-DSA-87 test vectors for its post-quantum signature precompiles
EIP-8355 adds ML-DSA-65 and ML-DSA-87 test vectors for its post-quantum signature precompiles — Coin Brief

What it means

Test vectors for the higher security levels are what client teams need before they can implement or compare the proposal. The open question for Ethereum is which of the two competing designs core developers choose, not whether a post-quantum verifier is needed.