On Tue, Nov 10, 2026 at 10:55:35PM -0500, Brian Sipos wrote:
> COSE WG,
> I'm running into a nuance of using COSE messages within a PKIX environment
> and maybe missing an existing explanation somewhere. Is there an explicit
> mapping to ensure interoperation between PKIX public key forms and
> equivalent COSE key sets or compatible COSE algorithm sets?

No. It is not guaranteed that algorithm mappings are unique. While I
think currently all mappings are in fact unique, the current plan for
PQC involves non-unique mappings.


> For example, the presence of pk-ec [1] in an end-entity certificate (with a
> specific EC curve) along with specific Key Usage extension of
> "digitalSignature" will constrain down to being compatible with a single
> one of the "ESP..." or "ESB..." COSE algorithm code points. Or for the
> other direction, given a specific COSE algorithm which public key form(s)
> are possibly compatible.

PK-EC keys can be mapped if the curve is supported in COSE.

There is no support for equivalent of KeyUsage. COSE key_ops is totally
broken (it comes from WebCrypto).


> The same kind of logic is needed to apply to COSE-using protocols such as
> LAKE/EDHOC but I can't seem to find an explicit interoperable mapping.
> Maybe it's so obvious that it's left unstated, but it seems like it's a
> complex enough topic that there will be edge cases and caveats.

I think the easiest way to deal with those in libraries is to have key
objects and allow loading those from X.509 stuff. Then have the
cryptographic operations use key objects.

That also precludes doing some really dumb stuff that has lead to
serious security issues.




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

Reply via email to