Signature algorithm and profile matrix¶
N-AALP's own cryptographic weight is a single object signature — the heavy ML-KEM key
exchange that secures a transport session is N-PAMP's concern and never lands on an N-AALP
object (the draft's Cryptographic Constructions section). Every N-AALP object is signed exactly once with COSE_Sign1 (RFC
9052), and algorithm agility is carried entirely in COSE's alg header parameter (the draft's Cryptographic Constructions section): changing the signature suite changes the alg value and the key type, with no change
to the COSE_Sign1 construction or the envelope. This page is the suite → algorithm → profile →
status matrix; it does not repeat the composite construction's byte-level detail, which lives
in crypto agility and migration.
Algorithm registry¶
| Algorithm | COSE alg |
Security level | Role | Source |
|---|---|---|---|---|
| ML-DSA-87 | -50 |
5 | post-quantum signature | FIPS 204; COSE identifier RFC 9964 |
| ML-DSA-65 | -49 |
3 | post-quantum signature | FIPS 204; COSE identifier RFC 9964 |
| ML-DSA-44 | -48 |
2 | post-quantum signature, edge/light tier only | FIPS 204; COSE identifier RFC 9964 |
| Ed25519 | -19 |
0 (classical) | opt-in composite leg only, never standalone | RFC 9864 §2.2 (also RFC 9053) |
| the opt-in composite | -65537 (provisional, private-use) |
3 (PQ leg) | ML-DSA-65 + Ed25519, LAMPS composite | N-AALP private-use allocation pending an IANA composite code point (the draft's Cryptographic Constructions section) |
These five rows are exactly the algorithm identifiers impl/go/cose/cose.go declares
(AlgMLDSA65, AlgMLDSA87, AlgEd25519,
AlgComposite65Ed25519 / AlgComposite44Ed25519)
and the security levels its algLevel function returns for each
(cose.go:60-75).
AlgComposite44Ed25519 (the edge-suite composite) is a reserved registry entry only — it is
registered but not implemented or graded this wave.
The profile table¶
One wire, one construction, one parameter row per profile (the draft's Cryptographic Constructions section). The profile is bound under the signature (envelope field 9 plus the protected header), so stripping or downgrading it invalidates the object:
| Property | Public (1) | Enterprise (2) | Sovereign (3) |
|---|---|---|---|
| Default signature | ML-DSA-65 (-49) |
ML-DSA-65 (-49) |
ML-DSA-87 (-50) |
| Minimum signature level accepted | 3 | 3 | 5 (refuses below) |
| Opt-in composite (Ed25519 + ML-DSA) | permitted | permitted | refused |
| Digest | SHA-384 | SHA-384 | SHA-384 or stronger |
| Composes with N-PAMP profile | Standard / High | High | Sovereign |
impl/go/cose/cose.go's
profileMinLevel
is the profile floor enforced at verify time: Sovereign requires level 5 and refuses anything
below (ErrProfileDowngrade); Public and Enterprise both require at least a level-3
post-quantum signature — no profile has a classical-only default. A Sovereign
verifier additionally refuses any composite (non-pure) object outright
(Verify1, cose.go:291-320,
ErrCompositeRefused), because CNSA 2.0 mandates ML-DSA-87 with no classical leg for
national-security signatures (the draft's Cryptographic Constructions section).
ML-DSA-44 (level 2, COSE -48) is licensed only in an explicitly negotiated edge/constrained
deployment outside these three editions — never as a silent downgrade of Public, Enterprise,
or Sovereign (the draft's Cryptographic Constructions section).
Status column¶
| Suite | Public | Enterprise | Sovereign | Status |
|---|---|---|---|---|
| ML-DSA-65 (pure) | default (MTI) | default (MTI) | refused (below floor) | implemented, graded |
| ML-DSA-87 (pure) | permitted | permitted | default (MTI) | implemented, graded |
| ML-DSA-44 (pure) | edge-only | edge-only | refused | registered; edge-tier only, outside the three editions |
Ed25519 + ML-DSA-65 composite (COMPSIG-MLDSA65-Ed25519-SHA512) |
permitted (opt-in) | permitted (opt-in) | refused | implemented, graded — the ten-port composite wave |
Ed25519 + ML-DSA-44 composite (COMPSIG-MLDSA44-Ed25519-SHA512) |
— | — | — | reserved (registered, not implemented) |
| SLH-DSA (FIPS 205) | — | — | — | reserved for the future (the draft's Cryptographic Constructions section) — a hash-based alternative to hedge the lattice monoculture, no envelope change needed to add it |
"MTI" (mandatory-to-implement) marks the suite every conformant implementation must support at that profile; "default" marks what a signer emits absent an explicit suite selection. Which suite signed a given object is carried under the signature itself as a signed suite declaration (naalp-object field 14, present iff the composite algorithm id is in use), so turning the composite on — or changing the default suite later — is a suite-id change through the registry, never a wire break (the draft's Cryptographic Constructions section).
The opt-in composite, briefly¶
The composite is one COSE_Sign1 whose signature value is the IETF LAMPS composite of an
ML-DSA-65 signature and an Ed25519 signature over a domain-separated message representative —
not two independent signatures (the draft's Cryptographic Constructions section). It exists for two reasons: bridging a
classical peer during migration, and as the pre-declared fallback path should ML-DSA ever be
weakened. impl/go/cose/cose.go implements both directions:
SignHybrid— the two-signature COSE_Sign (tag 98) form used for identity rotation objects (§5.2), not the object-signature composite.CompositeSigner.Sign/CompositeVerifier— the actual LAMPS composite object signature:ComputeMprimebuilds the domain-separated representative, the ML-DSA-65 leg signs it deterministically (rnd = 0) with context set to the suite label, and the Ed25519 leg signs it with no context.
Pure ML-DSA is SUF-CMA (strongly unforgeable); the composite is EUF-CMA but not SUF-CMA. This is never load-bearing for N-AALP's own security properties, because the single-use consume ledger keys off the content id and receipt chain, not raw signature bytes (the draft's Cryptographic Constructions section).
Digests¶
Every digest in the spine — content id, stream commitment, receipt chain — is SHA-384 or stronger, carried as a multihash so the algorithm is self-described and can be raised without a wire change (the draft's Cryptographic Constructions section). No profile admits a hash below SHA-384.
Signer identity depends on the algorithm¶
A signer's id is derived from its public key through a per-algorithm multicodec
(impl/go/identity/identity.go,
multicodecFor,
SignerID):
ML-DSA-87 uses multicodec 0x1212, ML-DSA-65 its own ML-DSA-65 multicodec, and Ed25519 0xed.
A composite signer gets a distinct derivation
(CompositeSignerID)
binding both component public keys, so a composite identity is never confusable with a pure
ML-DSA identity holding the same ML-DSA key. See identity and key lifecycle
for the full derivation and rotation model.
Failure modes¶
| Error kind | When |
|---|---|
ProfileDowngrade |
signature level is below the verifying profile's floor |
UnknownAlg |
the COSE alg value is not in the N-AALP registry |
HybridIncomplete |
the composite is present but either component signature fails to verify |
KeyAlgMismatch |
the key's algorithm does not match the object's declared header |
SuiteMismatch |
the signed suite declaration (field 14) disagrees with the actual signature structure or alg |
CompositeRefused |
a Sovereign verifier receives a composite (non-pure) object |
Every check is fail-closed (the draft's Security Considerations section): a failing object is rejected whole, returns its named error, and causes no state change.
What this page does not cover¶
- Per-language ML-DSA library sourcing. Unlike N-PAMP's KEM provider survey, this page does not audit which third-party library each of the ten SDKs uses to implement ML-DSA — that is a ten-port parity question tracked in the SDK build ledger, not a protocol-level algorithm choice. All ten ports are graded against the same non-circular oracle regardless of which underlying library each language binds.
- The composite's byte-level construction (the exact
M'formula, the LAMPS label octets, the concatenation order). That detail, including the primary-source re-verification date, is authoritative in crypto agility and migration and the draft's Cryptographic Constructions section.
See also: the object model, crypto agility and migration.