The term on this page
- ML-DSAModule-Lattice-based Digital Signature Algorithm
- the standardised post-quantum replacement for RSA and ECDSA signatures, published as FIPS 204
Also mentioned
RSARSARivest, Shamir and AdlemanA widely used public-key algorithm for encryption and digital signatures whose security relies on the difficulty of factoring large numbers.Read the full entry (new tab), ECDSAECDSAElliptic Curve Digital Signature AlgorithmElliptic Curve Digital Signature Algorithm, a widely deployed signature scheme based on elliptic-curve cryptography, offering strong security with compact keys.Read the full entry (new tab), TLSTLSTransport Layer SecurityTransport Layer Security, the protocol that encrypts and authenticates most internet traffic, including HTTPS. It uses key exchange, certificates, and symmetric encryption to protect a session.Read the full entry (new tab), FIPSFIPSFederal Information Processing StandardFederal Information Processing Standards, publicly announced standards developed by NIST for use in U.S. government computer systems, including cryptographic algorithms and modules.Read the full entry (new tab), NISTNISTNational Institute of Standards and TechnologyThe U.S. National Institute of Standards and Technology, the agency that develops and publishes cryptographic standards, including the FIPS series and post-quantum algorithms.Read the full entry (new tab), PQCPQCpost-quantum cryptographyCryptographic algorithms designed to run on today's classical computers while remaining secure against attacks by both classical and future quantum computers.Read the full entry (new tab), PKIPKIpublic key infrastructurePublic Key Infrastructure, the framework of certificate authorities, certificates, and policies that binds public keys to identities and enables trust in TLS, code signing, and email.Read the full entry (new tab) are defined in the glossary.
Read this first
RFC 10073 is now an IETF Proposed Standard for treating ML-DSA plus one traditional signature as a single X.509 algorithm. It defines 18 profiles, verifies both components, and recommends ML-DSA-65 with ECDSA P-256 for general use. It does not make old applications understand the new object identifiers, and it offers existential unforgeability rather than ML-DSA's stronger guarantee.
The IETF published RFC 10073 on 6 November 2026. It replaces the LAMPS working-group draft with a Standards Track specification for composite signatures in X.509 and related PKIX structures. Each composite combines ML-DSA with RSA, ECDSA, Ed25519 or Ed448 and exposes the pair as one algorithm identifier, one public key and one signature value.
The purpose is a transition hedge. A verifier accepts only when both component signatures are valid, so a break or implementation failure in one component does not immediately make the composite forgeable. That is useful for operators whose policy requires a post-quantum and a traditional component at the same time. It is not a reason to deploy all 18 choices.
What the standard changes
| Property | RFC 10073 rule | Operational consequence |
|---|---|---|
| Protocol shape | One composite algorithm identifier, public key and signature | Existing X.509 extension points can carry the composite without redesigning certificate syntax |
| Verification | ML-DSA and the traditional signature must both verify | Test failure of each component separately and reject any partial success |
| Profile choice | 18 registered combinations, none mandatory to implement | Choose and document a small profile instead of enabling every OID |
| Compatibility | Only upgraded software understands the new OIDs | Protocol compatibility is not application interoperability |
| Security target | EUF-CMA when at least one component remains secure and the prehash is collision resistant | Do not use it where strong unforgeability is required without a specific analysis |
The atomic shape matters. PKCS #10 requests, CMP, CMS, trust anchors and certificate processing can call the composite as if it were one signature algorithm. That is protocol compatibility. A legacy parser still sees an unknown OID and can reject the object, so the rollout still needs an explicit relying-party inventory and acceptance test.
Start with one profile, not eighteen
RFC 10073 registers combinations for three ML-DSA parameter sets and several traditional algorithms. Section 10.4 narrows the practical starting points:
| Constraint | Recommended composite | Maximum signature size |
|---|---|---|
| General use | ML-DSA-65 + ECDSA P-256 + SHA-512 | 3,381 bytes |
| RSA required | ML-DSA-65 + RSA-3072-PSS + SHA-512 | 3,693 bytes |
| Bandwidth sensitive | ML-DSA-44 + ECDSA P-256 + SHA-256 | 2,492 bytes |
| Level 5 required | ML-DSA-87 + ECDSA P-384 + SHA-512 | 4,731 bytes |
The general-use profile adds at most 72 signature bytes over pure ML-DSA-65, but size is only one constraint. Certificate chains add composite public keys and signatures at every affected level, and parsers, HSMs, enrollment services, revocation infrastructure and relying parties all need the same profile. Use the certificate and key format test guide to pin encodings before measuring the full chain.
The compatibility trap
The RFC is explicit that composite signatures do not provide application backwards compatibility. They use strict AND mode: an application must understand the composite OID and both components. A classical-only client cannot simply validate the ECDSA or RSA half, and an upgraded client must not accept only the ML-DSA half.
One object does not mean one migration step
A certificate can fit the existing X.509 structure and still fail across older enrollment clients, TLS stacks, document validators, HSMs or code-signing tools. Inventory every producer and verifier before changing the issuing profile.
Negotiated protocols may provide a separate compatibility path, but asynchronous artifacts do not. S/MIME messages, signed documents and code signatures can remain in circulation long after issuance, so a rollback plan must preserve verifiability of both old and new artifacts. Put that dependency map in the PQC migration roadmap, not in an undocumented certificate template.
The security limit is part of the decision
Pure ML-DSA targets strong unforgeability. RFC 10073 states that none of its composites is strongly unforgeable against a quantum adversary, and it does not recommend the construction where that distinction has not been shown to be acceptable. The composite can still provide existential unforgeability if at least one component remains secure and the selected prehash is collision resistant.
This is not a defect hidden in a footnote. It is a deliberate trade: the operator gets a hedge across two signature families and a form that fits existing X.509 interfaces, while accepting a narrower security property. Capture the application requirement before choosing the profile, especially for protocols that deduplicate signatures, treat signature bytes as identifiers or depend on non-malleability.
What PKI teams should test now
- Choose one profile based on policy and deployed traditional algorithms. Record the exact OID, prehash and component parameter sets.
- Generate a root, intermediate and leaf test chain. Verify key, signature and private-key encodings at every layer.
- Corrupt the ML-DSA component, then the traditional component. Both cases must fail without falling back to partial verification.
- Exercise certificate requests, issuance, renewal, revocation, chain building, export, import and backup. The ACME acceptance plan shows how to separate transport, account and certificate evidence.
- Measure full-chain bytes, parser limits, HSM calls, signing latency, verification latency and concurrent load. Do not extrapolate from a single leaf signature.
- Test every relying party and offline verifier that consumes the artifact. Unknown-OID failure is expected until that application is upgraded.
- Document the classical retirement condition, exception owner and rollback boundary. A hybrid without an exit condition becomes permanent complexity.
A scoped implementation audit can verify parsing, component failure and key handling. A migration engagement can sequence certificate authorities and relying parties, while conformance support can preserve the profile, vectors and negative tests as reviewable evidence.
What RFC 10073 does not do
It does not replace FIPS 204, make pure ML-DSA certificates obsolete, select one mandatory profile or certify an implementation under FIPS 140-3. It also does not make a composite certificate readable by existing applications. RFC 9881 remains the X.509 algorithm-identifier specification for pure ML-DSA, while RFC 10073 is the hybrid option for environments that have a reason to retain a traditional component.
The publication closes a standards gap, not an interoperability gap. The practical result depends on whether the complete certificate lifecycle and every verifier can process the same profile without weakening the two-component rule.
References
- IETF, RFC 10073: Composite ML-DSA for Use in X.509 Public Key Infrastructure, November 2026.
- NIST, FIPS 204: Module-Lattice-Based Digital Signature Standard, August 2024.
- IETF, RFC 9881: X.509 Algorithm Identifiers for ML-DSA, 2026.
- IETF, RFC 9794: Terminology for Post-Quantum Traditional Hybrid Schemes, June 2025.
- IETF LAMPS working group, composite ML-DSA test vectors, pinned by RFC 10073 to commit f0627ab.
Where this goes next