On Tue, Jun 17, 2025 at 08:28:48AM +0000, Lucas Prabel wrote:
> Hello all,
> 
> We have published a new version of 
> https://datatracker.ietf.org/doc/draft-prabel-jose-pq-composite-sigs/01/
> which describes how to use PQ/T composite keys and signatures for JOSE
> and COSE. We are pleased to welcome John Gray as a co-author of the
> draft, who will in particular help us align with the LAMPS composite
> draft when necessary.
> 
> The draft was presented at IETF 121, and numerous changes have been made 
> since.
> 
> In particular:
>  - as with draft-ietf-cose-dilithium-07, we now only support the seed
>    representation for the private key;
>  - we changed the composite signature combiner to align with the
>    upcoming version of the LAMPS composite draft;
>  - test vectors for JOSE have been added.
> 
> We welcome all comments, feedback and suggestions, and we plan to
> present the updated draft at IETF 123 in Madrid.

1) The domain separators being different makes this incompatible with
the LAMPS combiners.

This is a problem, because JOSE/COSE and X.509 would require
separate implementations of two algorithms that are different for no
apparent reason. And worse, that is crypto-grade code, which is not
cheap.


2) The key type is very badly named, and would not work with things
like ML-DSA+EdDSA (which are included in LAMPS draft for a reason).


What I think should be done is having this draft specify how to map
the signature construction in draft-ietf-lamps-pq-composite-sigs to JOSE
and COSE in the simplest way possible (in style of how section 6 of that
draft defines how to map the construct to X.509) and then define some
signature algorithms.

- SerializeSignatureValue gives byte string signature value, which is
  exactly what JOSE and COSE needs.

- The keys are algorithm-fixed, and SerializePublicKey and
  SerializePrivateKey gives byte string representations of public and
  private keys, which is exactly what AKP was designed for.

This way, one could use the same (higher-quality) implementation for
JOSE/COSE and X.509 with minimal differences.




-Ilari

_______________________________________________
jose mailing list -- [email protected]
To unsubscribe send an email to [email protected]

Reply via email to