 zeekay
|
df4a0ed733
|
magnetar Track B (research): no-reconstruct threshold SLH-DSA skeleton + proven FORS-leaf opening
Track B = research-only no-reconstruct threshold SLH-DSA (FIPS-205). FAIL CLOSED:
- threshold_noreconstruct.go: NoReconstructThresholdSigner iface; SignNoReconstruct
returns ErrNoReconstructUnproven; NoReconstructProven() hard-wired false;
AdmitMagnetarThresholdToPolarisMax() refuses Track B -> never silently admitted.
BurnLedger = one-time/few-time burn state (constraint 3); assertLeafWidth =
constraint-1 runtime guard (open path interpolates <= 1 leaf, never the seed).
- fors_threshold_open.go: distributed FORS-leaf threshold opening = the ONE
buildable, no-reconstruct, stock-FIPS-205-verifiable sub-protocol (approach B).
Leaf secrets shared directly (no master seed); combiner opens only md-selected
one-time leaves; assembled FORS sig byte-identical to centralized forsSign and
accepted by unmodified FIPS-205 Algorithm 17 (forsPkFromSig).
- Tests prove: byte-identity + stock verify (M192s/M256s, strict t-of-n),
one-time burn refusal, commit-mismatch rejection, fail-closed gates, leaf-width
guard, and a STRUCTURAL AST gate that OpenForsThreshold never calls a
seed-reconstruction primitive (contrast: existing strict-atom Combine does).
Finding: existing THBS-SE strict-atom Combine + dealerless PVSS-DKG both fully
reconstruct the master seed (derivedExpandInput / lagrangeScratch); the strict-atom
grep gate forbids only the variable NAME 'seed', not the reconstruction = Option A,
not strict no-reconstruct. Full Track-B signer stays fail-closed: hypertree
materialisation wall + keygen-MPC wall (documented in companion paper).
|
2026-06-28 01:05:48 -07:00 |
|