mirror of
https://github.com/luxfi/crypto.git
synced 2026-07-27 01:54:50 +00:00
Hanzo AI
1f07474b54
Adds NIST CAVS-style KAT regression for ML-DSA-44/65/87 and an
expanded-public-key form per EIP-8051 (~22,080-byte precomputed
A matrix for ML-DSA-65) for verifier-throughput experiments.
ethdilithium_compat subpackage re-implements the Keccak-substituted
verifier described in ZKNoxHQ/ETHDILITHIUM. NOT FIPS 204; for benchmark
and Ethereum-fallback only. Re-implemented from spec - no LGPL code
vendored.
The strict-PQ path remains crypto/pq/mldsa/mldsa{44,65,87}.{Sign,Verify}.
Patch-bump.
38 lines
1.6 KiB
Go
38 lines
1.6 KiB
Go
// Copyright (C) 2019-2025, Lux Industries Inc. All rights reserved.
|
|
// See the file LICENSE for licensing terms.
|
|
|
|
// Package kats implements Known-Answer-Test (KAT) regression
|
|
// vectors for the FIPS 204 ML-DSA-44 / ML-DSA-65 / ML-DSA-87
|
|
// parameter sets as wrapped by crypto/pq/mldsa.
|
|
//
|
|
// Each parameter set has a slice of pinned Vector records of the
|
|
// form (Seed, Msg, Ctx, PublicKey, PrivateKey, Signature). At test
|
|
// time, every vector is re-derived from its Seed via
|
|
// mldsaXX.NewKeyFromSeed and mldsaXX.Sign(..., randomized=false),
|
|
// then compared byte-for-byte against the pinned record. A
|
|
// mismatch is a regression in the deterministic-keygen or
|
|
// deterministic-sign path.
|
|
//
|
|
// # Origin
|
|
//
|
|
// FIPS 204 was finalized August 2024. NIST's ACVP-Server publishes
|
|
// a JSON test-vector battery (keyGen / sigGen / sigVer) that CIRCL,
|
|
// the upstream we delegate to, already runs against in its own
|
|
// test suite — those vectors are not redistributed here.
|
|
//
|
|
// The vectors in this package are SYNTHETIC: they are produced
|
|
// deterministically by the same code paths under test, pinned by
|
|
// hand, and serve as a regression tripwire — the test asserts that
|
|
// the bytes do not move silently between releases. Replace these
|
|
// vectors with official NIST CAVS .rsp files when those are
|
|
// published in a form suitable for direct ingestion.
|
|
//
|
|
// # Discipline
|
|
//
|
|
// The generator (kats/internal/gen) writes the same Go source the
|
|
// tests load. Running it twice produces byte-identical output, so a
|
|
// `go run ./pq/mldsa/kats/internal/gen` invocation that produces a
|
|
// diff is itself the regression signal. The Makefile target
|
|
// `make gen_kats` exposes this.
|
|
package kats
|