A malicious in-transit alteration seems like the hardest attack to defend. The expectation is that most in-transit alterations are innocuous changes made by a mailing list or spam filter, and that attackers can construct malicious original content more easily than they can construct malicious in-transit changes. My attempts to look behind an intermediary has focused on testing whether the message produced SPF Pass at origination, because the test can be applied to any address-server pair, and a local policy override can also be applied to any trusted address-server pair. Even if I can prove SPF Pass at origination, I would still be tricked by the malicious in-transit changes. But DKIM is also an insufficient defense, because (1) some messages do not originate with a DKIM signature, and (2) many domains have DMARC p=none or No Policy so we are reluctant to act on a broken signature.
Assuming that DKIM2 allows an evaluator to see the both before and after versions of a message, he will need a large language model to assess whether the "after" version is maliciously different than the "before" version. Additionally, we will always have traffic that does not have a DKIM2 signature. ARC allows the intermediary to make a statement about its business practices. DKIM2 allows the originator to make a statement about its messages. The two seem complementary to me. I guess the theory is that p=reject domains will embrace DKIM2, so the mailing list problem will be solved in a backhanded way: The mailing list problem will not be enforced against p=none domains, and DKIM2 will allow p=reject domains to participate. It does not really make the evaluator more secure, but it does solve the sender problem. On Tue, Apr 28, 2026 at 4:03 PM Richard Clayton <[email protected]> wrote: > -----BEGIN PGP SIGNED MESSAGE----- > Hash: SHA1 > > In message <[email protected] > il.com>, Douglas Foster <[email protected]> writes > > >Email authentication is only reliable when the message is received > directly > >from the originating organization. > > I don't think that's entirely true ... but let's move forward with that > > [snip] > > >When the limitations of DMARC are understood, I cannot construct a > scenario > >where ARC data causes harm. > > At $DAYJOB$ we implemented a scheme for using ARC to influence DMARC > decisions. It has no other real impact on "scoring" which is all about > email content and customer feedback. > > Using ARC information worked very badly because another large provider > failed to tell us that the contents of email had been altered (to make > it malicious) by a third party ... The third party resent the email > onward via the large provider who extended the ARC chain, so when we > walked it back to the beginning we saw that the first ARC set told us to > give the email a DMARC pass, which was an incorrect conclusion. > > The issue is that ARC envisages a chain of custody but the protocol > described in the RFC does nothing to check that chain doesn't include > entities that have changed the message without declaring they have done > so. So ARC data can cause harm if you think it is useful. > > When $DAYJOB$ realised that trusting the ARC chain to give us a DMARC > result was counter-productive we removed the other large provider from > our trust list. They are now, months later, stripping their own ARC info > from incoming email which may improve things. > > - -- > richard @ highwayman . com "Nothing seems the same > Still you never see the change from day to day > And no-one notices the customs slip away" > > -----BEGIN PGP SIGNATURE----- > Version: PGPsdk version 1.7.1 > > iQA/AwUBafESM2HfC/FfW545EQIH1QCfUg3hqAsXMGKtCOqn9nOZ4n1n3tAAnjmG > GMQJpGhqTs61IJrQFCCD+fsl > =UkSL > -----END PGP SIGNATURE----- > > _______________________________________________ > dmarc mailing list -- [email protected] > To unsubscribe send an email to [email protected] >
_______________________________________________ dmarc mailing list -- [email protected] To unsubscribe send an email to [email protected]
