On Sun, Oct 25, 2026 at 09:53:47PM -0400, John Levine wrote:
> >For this to work the DNS record needs to be modified to carry a
> >commitment to multiple key algorithms, the simplest approach is
> >to support multiple TXT records in the response RRset:
>
> We would really like to keep the DNS records for DKIM2 backward
> compatible and DKIM only has one key per selector. The current plan is
> for the DKIM2-Signature header to allow multiple selectors to do
> multiple signatures, but that still allows the rewrite using broken
> algorithms.
This is a hard choice. You can either have downgrade protection and a
new DNS record format, or a the legacy DNS format, and no downgrade
protection, you can't have both. The commitment to use strong keys
cannot arise from the presence of the strong signature in the message,
the attacker will strip that signature, and mutate the weak signature to
remove all trace of the sender supporting the second stronger signature.
> On the other hand, if the recipient only handles the broken one and
> hasn't been updated to handle a new strong one, we're still in the same
> place. I still think that we should say "don't use broken algorithms."
> It seems unlikely that DKIM will be a target of PQC unless the PQC is
> extremely cheap, so there's time to tell people to upgrade.
The downgrade attack is effective against recipients that support strong
keys, by stripping the strong signatures. Recipients will of course
still support the weaker keys for some time, ... because interoperability.
--
Viktor. 🇺🇦 Слава Україні!
_______________________________________________
Ietf-dkim mailing list -- [email protected]
To unsubscribe send an email to [email protected]