[Ietf-dkim] Re: New I-D: A Deployment Profile for DKIM2 via Milter Interface

[Bron Gondwana <[e-mail]>]

On Sun, Apr 19, 2026, at 22:37, Vittorio wrote:
> Hello Bron,
> 
> bad guys lie about recipes too. Multiple colluding nodes forge recipes 
> with the same keys they use to forge bh= values. If the adversary 
> controls N nodes in the chain, those N nodes produce N internally 
> consistent but fabricated recipes... and the receiver has no way to 
> distinguish them from honest ones.
> "Bad guys lie" is not an argument for body recipes. It is an argument 
> against trusting any single mechanism (including body recipes).

I don't see how that is possible. Perhaps you can generate an example to demonstrate how nodes can fabricate recipes that don't reliably reproduce the content of the previous hop and yet validate against the previous hop's bh=.

My contention is that with a recipe and a hash; it is impossible (short of breaking the hash function) to lie about which changes you made.  Sure, you can lie about anything that happened inside a range of domains that you control, since you can change and re-hash the message and lie about the individual hops that happened inside that range, but you can't make any change appear to have been made by another system which is NOT one of those domains you control.

Bron.

--
  Bron Gondwana, CEO, Fastmail Pty Ltd / Fastmail US LLC
  [e-mail]