Webinar Nov. 24: Governing Crypto-Agility Register

Hybrid signatures: what changes for PKI?

Hybrid Signatures for Certificates

The IETF has published RFC 10073 and RFC 10082, which standardize composite ML-DSA signatures for X.509 certificates and for CMS. They give PKI teams an interoperable way to apply the hybrid approach that most European cybersecurity agencies recommend for the post-quantum transition.

Which uses do the two RFCs cover?

RFC 10073 defines how ML-DSA, the post-quantum signature algorithm standardized by NIST in FIPS 204, is combined with a traditional signature algorithm (RSA, ECDSA, Ed25519 or Ed448) to form a single “composite” algorithm for X.509 public key infrastructure.

RFC 10082 specifies how the same composite algorithms are used in the Cryptographic Message Syntax (CMS), the format that underpins S/MIME email, signed documents and many code and firmware signing schemes. Both documents were developed in the IETF LAMPS working group and published on the standards track.

Why combine two algorithms instead of one?

RFC 10073 sets out two reasons to combine a post-quantum algorithm with a traditional one.

The first is dual-algorithm protection. An attacker must break both components to forge a composite signature, which hedges against weaknesses that may still be found in ML-DSA or in specific implementations of it. ML-DSA is recent, and its implementations have far less field experience than those of RSA or elliptic curves.

The second is flexibility during the transition. An organization that already runs a mature, validated and certified implementation of RSA or ECC can add an ML-DSA component next to it. Long-lived signatures gain post-quantum protection immediately, while the new implementation matures and goes through certification.

This matches the position of most European agencies. In a joint statement published in June 2025, cybersecurity agencies from 21 European states, including ANSSI and BSI, strongly recommended deploying post-quantum cryptography in hybrid solutions. ANSSI also requires hybridization for products that include post-quantum mitigation and seek a security visa, with hash-based signatures as the exception. Other agencies take a different view: the NSA does not require hybrid schemes for US National Security Systems, and the UK NCSC treats them as an interim measure. For organizations that follow European guidance, hybrid remains the reference path.

Until now, that recommendation lacked a standardized mechanism for certificates. In its February 2026 guides on the post-quantum transition of IPsec, SSH and TLS 1.3, ANSSI noted that no official standard yet existed for hybrid authentication based on signatures and certificates. RFC 10073 and RFC 10082 fill that gap for X.509 and CMS.

How does a composite stay compatible with existing protocols?

A composite key holds two key pairs, one ML-DSA and one traditional, generated independently. Signing produces both signatures over the same message representative, and verification succeeds only if both are valid.

Seen from the outside, the composite has a single public key, a single signature value and a single algorithm identifier. The RFC calls this property protocol backwards compatibility: X.509, PKCS#10 certificate requests, CMP and CMS can process a composite like any other signature algorithm, through the algorithm agility mechanisms they already contain. These protocols need no modification to carry hybrid signatures, because the complexity is handled inside the cryptographic library or module.

RFC 10073 registers eighteen combinations, pairing the three ML-DSA security levels with RSA-PSS, RSA PKCS#1 v1.5, ECDSA on NIST and Brainpool curves, Ed25519 and Ed448. The breadth is deliberate, so that any algorithm deployed today can serve as a starting point. In practice, most organizations will use one or two combinations, and the RFC provides a short list of recommended choices for common use cases.

Where are the deployment pitfalls?

  • Composites operate in a strict “and” mode. A relying party that has not been upgraded cannot verify a composite signature, so the order in which issuing and verifying systems are upgraded has to be planned.
  • Composite ML-DSA provides existential unforgeability (EUF-CMA) as long as at least one component remains secure, but it does not offer strong unforgeability (SUF-CMA) against a quantum adversary. The RFC advises against its use in applications where EUF-CMA has not been shown to be sufficient, which calls for a review of each use case.
  • Component keys must never be reused, whether in another composite or as a standalone key.
  • Sizes increase. An ML-DSA-65 public key is 1,952 bytes and its signature 3,309 bytes, before adding the traditional component. Certificate chains, constrained devices and protocols with size limits deserve early testing.

Do independent implementations agree?

A standard only delivers value if independent implementations agree on it. Within the IETF Hackathon PQC Certificates group, implementers publish certificates, keys and signed messages in a shared repository and verify each other’s outputs, with results tables made public after each round.

Composite formats have therefore been exercised across multiple implementations well before publication. Organizations selecting PKI or signing products can usefully ask vendors for their interoperability results against the final RFC versions.

Beyond signatures, how far will the composite approach extend?

Two companion documents, defining composite ML-KEM for X.509 and for CMS, are nearing completion in the same working group. Work on composite ML-DSA for TLS authentication has also been adopted by the TLS working group. Taken together, these documents extend the composite approach from signatures to key establishment, and from certificates and messages to secure channels.

CryptoNext’s contribution

Daniel Van Geest, Senior Research and Development Engineer at CryptoNext Security, is a co-author of RFC 10082 and contributed reviews, text and interoperability testing to RFC 10073. He is also a co-author of the composite ML-KEM document for CMS and a key contributor to its X.509 counterpart.

CryptoNext engineers are authors of RFC 9802, RFC 9882, RFC 9909 and RFC 9936, and of working group documents on CMS SignedData under EUF-CMA signatures, composite ML-DSA in TLS and FN-DSA in CMS. CryptoNext implementations take part in the IETF Hackathon PQC Certificates interoperability testing.

Daniel Van Geest sums up what the RFCs bring to PKI teams:

“European agencies have been recommending hybrid post-quantum signatures for years, but PKI teams had no standard way to put them into a certificate. With these RFCs, a composite is just another signature algorithm to X.509 and CMS. The hard work moves into the cryptographic library, which is exactly where it belongs.”

He also adds a word of caution about the transition:

“Composites give organizations a way to start protecting long-lived signatures now, without betting everything on algorithms that are still young. They are not a drop-in fix, though. Any system that verifies signatures has to be upgraded first, and that starts with knowing where those systems are.”


Would you like more information about our company, its products, our partner programs or job opportunities?

To address quantum risk, embrace crypto-agility