mirror of https://github.com/luxfi/threshold.git synced 2026-10-30 15:15:53 +00:00

Files

T

zeekay 7af0617109 docs: rip AI slop in threshold/study (magnetar.md 243→140, cross-family-defense 93→75, README 85→74)

Patterns ripped: profile-name-selection prose, honest-accounting framing,
ships-today/research-track dichotomy, checklist-restates-prose bullets.
Zero spec loss.

2026-05-18 21:21:12 -07:00

3.2 KiB

Raw Permalink Blame History

Cross-family defense in depth

Why Aurora (Pulsar ‖ Corona) and Nova (Pulsar ‖ Corona ‖ Magnetar) name distinct profiles, and why the third leg matters.

Lattice-family inclusion

Module-LWE with module rank k=1 is Ring-LWE. There are polynomial-time reductions between them with parameter shifts in modulus and error width. Stacking Pulsar (MLWE) and Corona (RLWE) gives:

Implementation-level diversity (different code, parameter sets, rejection samplers, DKG ceremonies).
NOT family-disjoint hardness — a structural break against lattice cryptography compromises both.

Aurora is named after a single-cause event, not a redundancy construction: both lattice-family events that could trigger it share a common cause.

What hash adds

Magnetar (FIPS 205 / SLH-DSA, public-DKG + MPC) rests on hash collision and preimage resistance, not on any lattice assumption. A polynomial-time MLWE oracle does not yield a hash-collision oracle, and vice-versa. Cross-family disjointness:

MLWE break → compromises Pulsar AND Corona.
Hash break → compromises Magnetar AND every hash-anchored primitive in the stack (SHA-3, BLAKE3, Merkle trees, …) — but Pulsar and Corona survive.

Either failure leaves at least one of the three primitives standing. Nova safety property.

Production stages

Stage A — GPU batch verify of single-party SLH-DSA (live). Nova gates on a batch of per-validator SLH-DSA sigs verified through one LatticeOps.SLHDSAVerifyBatch dispatch. Throughput bounded by N · (single-party SLH-DSA verify), amortised by GPU.
Stage B — Pedersen-DKG + MPC threshold SLH-DSA (slot 0x012207 in LP-0120, in cryptographic review). Nova collapses to a single threshold-SLH-DSA cert. Lineage: Goyal–Kothapalli–Masny–Mukherjee IACR 2024/447 + Pedersen-VSS auditability layer.

A code-family threshold scheme (HQC, NIST PQC4 backup KEM) is research-track only; HQC is a KEM, code-based signature schemes (CFS, SDP-based) are not NIST-tier. Magnetar is the third leg; Nova safety holds whether Stage A or Stage B is active.

Wire impact

Profile         Construction                       Per-block wire
──────────      ─────────────────────────────────  ──────────────
Pulsar          Module-LWE threshold (floor)         ~3.3 KB
Aurora          Pulsar ‖ Corona                      ~36 KB
Nova (Stage A)  Pulsar ‖ Corona ‖ Magnetar           ~36 KB + N × 35.7 KB
                  (N = signing committee, batched verify)
Nova (Stage B)  Pulsar ‖ Corona ‖ Magnetar           ~36 KB + 35.7 KB
                  (single MPC threshold SLH-DSA, slot 0x012207)

Stage A is heavy (third leg = N per-validator SLH-DSA sigs batched at verify time). Stage B brings it down to a single threshold-SLH-DSA cert; verifier ABI unchanged (standard FIPS 205 verifier).

3.2 KiB Raw Permalink Blame History Unescape Escape