Consider using COSE algorithm identifiers for credential_signing_alg_values_supported for mso_mdoc

[awoie]

ISO mdocs use COSE_Sign1 to secure the MSO using COSE algorithms, not JOSE algorithms. COSE algorithms are identified by integer values according to the IANA registry. We should change the mso_mdoc section to use those identifiers instead of JOSE algorithm identifiers.

[awoie]

However, when using COSE algorithms, the issue is that some algorithm identifiers such as ES256 (-7) are polymorphic identifiers. For example, it is possible to use any brainpool curve, or NIST curve (of any length) with ES256 since ES256 only refers to ECDSA with SHA256. I know that COSE is going to deprecate polymorphic identifiers but 1) ISO will keep referring to ES256 (-7), and 2) fully defined algorithm identifiers have not been registered with IANA yet.

As an interim solution, we could say that credential_signing_alg_values_supported refer to JOSE algorithms which should then be mapped to their COSE counter part. Currently, the spec says to use one of the algorithms defined in ISO but ISO does not have a list of algorithm identifiers.

[awoie]

Currently, we have:

Cryptographic algorithm names used in the credential_signing_alg_values_supported parameter SHOULD be one of those defined in [ISO.18013-5].

We would need to change to:

Cryptographic algorithm names used in the credential_signing_alg_values_supported parameter SHOULD be one of those defined in [ISO.18013-5] represented as JOSE algorithm identifiers.

[Sakurann]

We would need to change to:

Cryptographic algorithm names used in the credential_signing_alg_values_supported parameter SHOULD be one of those defined in [ISO.18013-5] represented as JOSE algorithm identifiers.

Sounds like we would also need a sentence saying that JOSE algorithms for credential_signing_alg_values_supported should then be mapped to their COSE counter part.

(the proposal sounds a little bit counter intuitive, but guess 1) ISO will keep referring to ES256 (-7), and 2) fully defined algorithm identifiers have not been registered with IANA yet are the arguments why using JOSE and then mapping is better...)

I think this is ready for PR.

[Sakurann]

ping @davidz25 @GarethCOliver @hlozi

[awoie]

This is a potential duplicate of #354. I believe to support this, additionally we would need to introduce an optional credential_signing_crv_values_supported parameter that can be used to indicate the curves supported since this would be needed for some of the polymorphic signature algorithms, e.g., ES256 in COSE.

[Sakurann]

WG discussion

• VP and VCI should have a similar mechanism for this
• match what we did in clarify issuer_signed_alg_values and device_signed_alg_values OpenID4VP#553