Files
Antje Worring df4c537c16 magnetar: remove proof-by-rename + circular proofs; honest threshold status
True no-reconstruction threshold SLH-DSA is impossible in the no-dealer /
no-preprocessing model a public, leaderless, permissionless chain needs
(Kondi-Kumar-Vanegas: extractable hash-based signatures cannot be thresholded
by black-box hash use). Stop pretending otherwise.

De-cheat:
  - DELETE the circular/vacuous proofs that manufactured false assurance:
    Magnetar_N1_Atom_Refinement.ec, Magnetar_N1_SHAKE_Expand.ec,
    Magnetar_N4_KeyDeriveStable.ec, lemmas/{Magnetar_CT,SLHDSA_Functional}.ec.
    The headline "strict-atom byte-equality" theorem was `apply <axiom that
    restates the theorem>`; the Lean side was `sorry`/`:= True`.
  - Re-open MAGNETAR-STRICT-ATOM (BLOCKERS): the public combiner DOES
    reconstruct the full FIPS 205 master every signature; the v1.1 "closure"
    only renamed identifiers. PROOF-CLAIMS / AXIOM-INVENTORY / TCB docs now say
    what the code does.
  - Re-label the name-grep "strict-atom" / "CT" checks as identifier-hygiene
    lint, NOT security or constant-time properties.
  - PVSS-DKG open-reveal (publishes the master to any observer) is gated as a
    TEST-ONLY path with HONEST LIMITATIONS; production does not rely on it.
  - Remove dead htRootCompute; staticcheck clean.

Honest three-leg posture (SPEC 1.0, BLOCKERS):
  - Permissionless production = INDEPENDENT FIPS 205 sigs + the weighted quorum
    certificate (luxfi/consensus), optionally STARK/FRI-compressed (luxfi/p3q).
    No key sharing, no reconstruction.
  - Trusted-hardware custody = TEE-attested combiner (trust-relocation, NOT MPC).
  - THBS-SE = RESEARCH-ONLY (transient seed reconstruction at the combiner);
    the T-SLH-DSA-MPC track (MPC over SHAKE) is the other research escape hatch.

Tests green (CGO=1, 71s). Net -923 lines.
2026-06-21 13:06:57 -07:00

123 lines
5.2 KiB
Markdown

# Magnetar --- design
Magnetar is the hash-based (FIPS 205 SLH-DSA) PQ leg of the Lux
cross-family signature suite. This document states the architecture and
why each path exists.
## The fundamental constraint
SLH-DSA is a tree of hash computations: the secret is a SHAKE seed and
signing traverses a hypertree of one-time / few-time signatures keyed
by that seed. There is NO linear-aggregation identity (unlike
FROST/Schnorr or the Raccoon-style ML-DSA threshold). The literature is
explicit:
- Cozzo & Smart, EUROCRYPT 2019, *Sharing the LUOV*: every internal
SHAKE/SHA-2 evaluation is a non-linear function of the secret seed.
- Bonte, Smart & Tan, 2023, *Threshold SPHINCS+*: there is no efficient
t-of-n SLH-DSA scheme producing one FIPS 205-shaped signature without
either (a) reconstructing the seed in some combiner process, or (b)
full MPC over the SHAKE hash tree (multi-second, multi-megabyte per
signature).
- NIST IR 8214: SLH-DSA is the highest threshold-MPC cost among FIPS
20{3,4,5}.
True no-reconstruction threshold SLH-DSA is OPEN RESEARCH. Magnetar
does not pretend otherwise. Instead it offers three paths chosen by
deployment model.
## The three paths
### 1. Per-validator standalone --- PRODUCTION DEFAULT (sound)
`standalone.go`. Each validator holds its own FIPS 205 keypair, signs
the message independently, and the consensus layer collects N
single-party signatures into a `ValidatorAggregateCert`. Verification
counts valid signatures against the validator-set registry; the quorum
decision lives at the consumer.
This is the sound primitive: there is no shared seed, no DKG, no
dealer, and no reconstruction anywhere. It sidesteps both (a) and (b)
by emitting N independent signatures. Wire size is N x |sigma| (a
Z-Chain Groth16 rollup compresses it; separate primitive).
THIS IS THE DOCUMENTED PRODUCTION PRIMITIVE for public-BFT post-quantum
consensus.
### 2. TEE-attested combiner pool --- OPT-IN PRODUCTION (sound under attested-hardware trust)
`luxfi/threshold/protocols/slhdsa-tee`. For deployments that genuinely
need a SINGLE FIPS 205-shaped signature from a committee (custody,
bridge signing), Magnetar routes Combine through a t-of-n
attested-combiner pool, gated by the single profile function
`magnetarRefuseUnderStrictPQ`.
Honest trust model: the seed is reconstructed in full, but INSIDE a
measured enclave whose binary + firmware chain-validate to a vendor
root (AMD KDS today; Intel TDX / NVIDIA NRAS are stubs). This is
TRUST-RELOCATION --- it moves the reconstruction into attested hardware
and ADDS that hardware (plus the attestation chain and the t hosts) to
the TCB. It is NOT MPC and does NOT keep the seed secret from every
party; the enclave sees it.
### 3. Permissionless THBS-SE --- RESEARCH-ONLY
`thbsse.go` + `thbsse_field.go` + `thbsse_assemble.go` +
`slhdsa_internal.go`. A t-of-n committee, slot-bound commit-and-reveal,
and an anyone-can-combine public combiner. The public combiner
RECONSTRUCTS the full FIPS 205 master into one buffer every signature
(`ASSEMBLE-INVARIANT.md`) and emits a byte-identical FIPS 205
signature.
Because whoever runs Combine sees the seed, there is effectively a host
in the TCB at sign time. This path is RESEARCH-GRADE. It must not be
presented as production or as "no host in the TCB". The earlier "strict
-atom Combine closed the no-leak gap" claim was proof by rename (it
avoids the variable name `seed`, not the reconstruction). See
`BLOCKERS.md::MAGNETAR-STRICT-ATOM` (OPEN).
## Setup: dealer, dealerless, or TEE
- **Dealer (`NewThbsSeKey`)**: one machine samples the master,
byte-shares it over GF(257), and zeroizes it. The dealer is in the
TCB for setup only. KAT-reproducible; recommended for foundation-HSM
bootstrap.
- **Dealerless production (`RunDKG`)**: per-party PVSS with hash
commitments; the transcript carries NO constant-term reveals, so the
master is not reconstructible from the wire. BUT pk derivation
reconstructs the master transiently (inherent: `pk = SLH-DSA.PK(M)`
is a hash tree over M), and the path does NOT robustly exclude
malicious dealers at DKG time (hash commitments aren't openable
without revealing m_i). See `RunDKG` HONEST LIMITATIONS in
`pvss_dkg.go`.
- **TEE**: routes both pk derivation and Combine through attested
hardware (trust-relocated).
A leaking open-reveal DKG (`RunDKGSimulationOpenRevealTestOnly`) exists
for KAT replay only, behind an explicit hazard acknowledgement; its
transcript reveals the master.
## Parameter sets (FIPS 205 SHAKE)
| Mode | NIST Category | sig bytes |
|---|---|---|
| SHAKE-192s | 3 | 16224 (recommended) |
| SHAKE-192f | 3 fast | larger sig, faster sign |
| SHAKE-256s | 5 | 29792 |
SLH-DSA signatures are large (16-30 KB) vs ML-DSA (2-5 KB); the
per-signature cost dominates economics, which is another reason the
standalone path (no MPC ceremony) is the production default.
## What is NOT claimed
- No mechanized proof of any threshold property (the prior EC/Lean
track was vacuous; see `PROOF-CLAIMS.md`).
- No no-leak property for THBS-SE.
- No robust-against-malicious-dealers property for the production DKG.
- No external audit.
Use the per-validator standalone path for production public-BFT
signing. Use the TEE pool when you are willing to trust attested
hardware. Treat THBS-SE as research.