Inline:

On Wed, Jul 31, 2024 at 1:27 PM Ilari Liusvaara <[email protected]>
wrote:

> On Wed, Jul 31, 2024 at 11:03:47AM -0500, Orie Steele wrote:
> > key_ops is defined to parallel to JWK and web crypto key usage:
> >
> > https://www.rfc-editor.org/rfc/rfc9052.html#section-7.1
>
> I think key_ops is just FUBAR.
>
>
> > I think you are asking about derive key (7) and derive bits (8).
> >
> > The same questions came up with COSE / JOSE HPKE.
> >
> > Which of these operations corresponds to KEMs?... should HPKE use
> "key_ops"
> > that mix both kem and symmetric key operations? ( "derive key" +
> "encrypt"
> > ) ?
> >
> > AFAIK, there is no guidance on this, and no registry of key operations
> > which is extensible (which could be a good thing).
>
> I think it is whatever permission is required in webcrypto to implement
> HPKE. Which does not make any sense any other way.
>

Then I think it's deriveBits for HPKE... although when importing keys for
use with HPKE I think it's common to delete "key_ops", or set it to be the
empty array.
Should an application throw an error before doing this if it encountered
"key_ops" that it did not expect for HPKE? what should it expect for HPKE
or edhoc?


>
>
> > The pop use case is a bit more specific, because pop usually means some
> > form of nonce signing.
>
> Key_ops works for signing, but not for verification.
>

afaik, this would be legal in JWKs used for pop:

public key ... "key_ops" "verify", "use" "sig",
private key ... "key_ops" "sign"


>
> > I have wondered if JOSE and COSE would benefit from "more fine grained"
> key
> > operation restrictions.
>
> JOSE has "use", which is quite coarse-grained but sensible.
> Unfortunately, COSE does not have equivalent.
>

Yep, it's "sig" or "enc"... that's pretty coarse.

https://datatracker.ietf.org/doc/html/rfc7517#section-4.2


> And more fine-grained restrictions would make sense. E.g., restricting
> key to a single family of properly-separated operations (which is
> cryptographically kosher thing to do).
>
>
> > I think a new concept that is not tied to web crypto might yield better
> > security properties in the long run.
>
> I think it would lead to better security properties.
>

I hope this was not a typo and that we agree : )


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


-- 


ORIE STEELE
Chief Technology Officer
www.transmute.industries

<https://transmute.industries>
_______________________________________________
COSE mailing list -- [email protected]
To unsubscribe send an email to [email protected]

Reply via email to