Hello COSE group, when this document was presented in the last IETF meetings, there was an open question on interest from the group. I participate rarely in WG activities, but I use COSE a lot in OSCORE and LAKE, and
I would appreciate having this taken up the group.
(In fact I plan to contribute LAKE cipher suites to it, and other glue
where necessary).
Some notes on the document from reading it in full (some with text
proposals in [PR2]) -- I hope that beyond improving the document, they
also illustrate its maturity:
* [PR2] Abstract: At this point in time I think it suffices to say the outcome
(rather than the history) of the NIST processes, and provide more
information about which parts of it are made available to COSE
(currently "authenticated encryption, hashing, Message Authentication
Code (MAC), and related operations", but maybe spelling out HKDF and
HMAC makes sense).
* The updates from the top half of table 8 might better be done in prose
(it's hard for a reader to compare the table, and AIU the change is
mostly just "Add *+HKDF-Ascon-Hash256 where there are other entries
already").
On the 2nd half: Why are those entries needed here and not with
others? (Or should they be the same as the already existing entries
-3..-1, and just extended like the rest of the table?
(Same applies to table 12).
* On 3.2.2+ vs 3.3, readers will appreciate some guidance on whether to
use Hash plus the Hash-to-X adapters HKDF and HMAC, or whether to use
Ascon-MAC directly.
My intuition would be that using the Ascon-MAC versions would be
prefereable (more efficient), but if a protocol prescribes using HKDF
(e.g. to make use of a split between Extract and Expand), going
through HKDF might be necessary.
The outcome of such recommendations might also be reflected in
requested algorithm numbers: Users in CBOR-related protocols have a
tendency to infer recommendations from encoding lengths, so if the
general recommendation is to only use Ascon-Hash for MAC and KDF in
exotic situations, it might make sense to request larger values for
them.
(Note mostly to self: LAKE (RFC9528) almost prescribes HKDF, but
actually just for SHA-2, and has explicit instructions for other hash
functions, which will make extending it funny).
* Appendix A: It will be helpful for users of the final document to have
a test vector for each of the algorithms; possibly, alike vectors can
reuse most of their lines.
* [PR2] Some editorial nits not worth spelling out in detail.
BR
Christian
[PR2]: https://github.com/dmochkas/draft-ochkas-cose-ascon/pull/2
--
I shouldn't have written all those tank programs.
-- Kevin Flynn
signature.asc
Description: PGP signature
_______________________________________________ COSE mailing list -- [email protected] To unsubscribe send an email to [email protected]
