Re: [dmarc-ietf] easier DKIM, DMARC2 & SPF Dependency Removal

[Douglas Foster <[e-mail]>]

We are talking about SPF AND DKIM because of the problems with DKIM
replay.   Can someone summarize the state of the DKIM update options that
have been ruled in or ruled out?

We have the DARA proposal from Google, which is related to signature
replay, but focused on ARC.  I am guessing that it correlates closely to
the DKIM updates.  Should the discussion move to that draft for a bit?

I also look forward to Scott's ideas.

DARA limits replay by linking the signature to a recipient.   An
alternative approach is to limit replay be linking the signature to a
source server.   It may be less effective, but some benefit can be gained
from this approach, which does not require a DKIM rewrite.   Here are my
thoughts:

Given a received sequence of

   - Msg Date
   - Rcv Date 1
   - Rcv Date 2
   - ....
   - Rcv Date N


A signature should have a start time between one of these date boundaries.

   - If the signature start time is equal to, or shortly, after, a Received
   header timestamp, the signing server is the BY server which create the
   Received header.
   - If the signature start time is prior to any Received header timestamp,
   but on or after the message Date header, then the FROM information on the
   first Received header identifies the signing server.
   - If the signature start time is prior to the message Date header, the
   signature looks suspicious

In some cases, clock skew may create ambiguity about the signing server.
Even when this occurs, the ambiguous servers are likely to lie behind the
same perimeter server, so the ambiguity may not need to be resolved.

The first global IP after the signature server becomes the perimeter server
which must demonstrate that it's domain has the right to act on behalf of
the signing domain.

   - We do an SPF test on the perimeter server's IP address and the
   signature domain.  If this test produces SPF PASS, the signature has
   enhanced trust.
   - If the perimeter server domain matches the signature domain, and the
   perimeter server host name can be forward confirmed to its IP address, this
   provides an alternative to SPF PASS.
   - If either test succeeds, the risk of DKIM reply is significantly
   lessened.   If neither test succeeds, then the signature risk of DKIM
   replay is greater

Improvements within control of the signing server:

If the message is not created by the originating server, the message should
contain one or more Received headers.   The header list on the signature
should contain "Received" entries to match the number of Received headers
at the time of signing.   This further identifies the signature's position
in the Received chain.

Improvements requiring changes to the DKIM specification

They could identify a way to document the signing server's domain in the
signature.   Then the perimeter server host names can be tested to see if
they match the signature's server domain and can be forward-confirmed to
the Source IP.   This provides an alternative to SPF for configurations
where the server domain and the signature domain are different
organizations.

Doug Foster



On Wed, Jun 28, 2023 at 3:44 PM Scott Kitterman <[e-mail]>
wrote:

> I think it's quite relevant.
>
> The assumption that this is based on is there's a need to specify because
> SPF data is not reliable enough for everyone.  If that premise is correct
> (I don't agree with it, but it's a separate issue), then I think telling
> people that they should use DKIM because it IS reliable, when it's got its
> own issues isn't a great idea.
>
> I've been mulling this whole topic over and I think I'm close to having
> mulled it enough to have a useful proposal.  SPF bad, DKIM good is a gross
> over-simplification, but so is if it passed SPF, it's authorized, so go
> whine at the provider.
>
> Scott K
>
> On June 28, 2023 6:32:41 PM UTC, Barry Leiba <[e-mail]>
> wrote:
> >I think DKIM replay is largely irrelevant to this discussion (about
> >the sender specifying which authentication to use), because if that's
> >your biggest concern with respect to DMARC, then "SPF only" is the
> >answer.  "SPF *and* DKIM" is not any better than that.
> >
> >> You seem to imply that auth=dkim+spf wouldn't be effective against DKIM
> reply
> >
> >(Assuming you mean "replay".)  "SPF and DKIM" does not give any
> >benefit beyond "SPF only" in this case.
> >
> >Look, either SPF fails because the message was relayed illegitimately...
> >...or SPF passes because the replayer used the sender's legitimate
> >infrastructure to do the replay.
> >
> >Barry
> >
> >On Wed, Jun 28, 2023 at 12:43 PM Alessandro Vesely <[e-mail]>
> wrote:
> >>
> >> Thank you for your analysis.  However, it doesn't touch on DKIM replay.
> >>
> >> I know this topic belongs to the other list.  Let me briefly recall it,
> if this
> >> doesn't take too many cycles from core matters:  It occurs when a signed
> >> message is replayed by unauthorized hosts to recipients which were not
> >> originally addressed.  So, it is one case of your 3rd proposition: In
> some
> >> scenarios, DKIM will pass when SPF fails.
> >>
> >> You say that it is technically unnecessary to test both because DKIM
> should
> >> always pass when SPF passes (1st proposition).
> >>
> >> You claim:
> >> > But where the harm comes is in cases of mis-configuration, because
> now if
> >> > *either* of them is misconfigured, the whole thing fails -- neither
> of them
> >> > serves as a backup for the other; instead, the misconfiguration of
> either
> >> > one causes deliverability problems.
> >>
> >>
> >> I agree.  But what if SPF and DKIM are both configured properly?  You
> seem to
> >> imply that auth=dkim+spf wouldn't be effective against DKIM reply, but
> your
> >> analysis doesn't cover that case explicitly.
> >>
> >> Perhaps there are better ways to counter that specific problem, and
> certainly
> >> it's not what this WG is tasked to do.  But, just to make the point, I
> think
> >> it's interesting to know.
> >>
> >>
> >> Best
> >> Ale
> >>
> >>
> >> On Tue 27/Jun/2023 16:24:21 +0200 Barry Leiba wrote:
> >> > I don't understand how most of your message fits into this discussion:
> >> > you're comparing SPF's policy points with DMARC policy.  we're talking
> >> > about SPF as an authentication mechanism together with DKIM (not
> >> > DMARC) as an authentication mechanism... and then using those
> >> > authentication results in DMARC policy evaluation.
> >> >
> >> > But here: I've said all this before in separate places, so I'll put it
> >> > in one place, here, one more time:
> >> >
> >> > Given that SPF and DKIM are both configured properly:
> >> > 1. If SPF passes, DKIM will always pass.
> >> > 2. If DKIM fails, SPF will always fail.
> >> > 3. In some scenarios, DKIM will pass when SPF fails.
> >> >
> >> > Therefore, when everything is configured properly, SPF adds no value
> >> > beyond what DKIM does, and DKIM does add value beyond what SPF does.
> >> > That's why I am (and others are) arguing that we should remove SPF
> >> > *from DMARC evaluation*.  There's no argument that for now, or some,
> >> > SPF outside of DMARC still has value.
> >> >
> >> > What others are arguing is that in the real world, things do get
> >> > mis-configured, and if DKIM is misconfigured and fails when it
> >> > shouldn't, SPF adds value by providing a working authentication.
> >> > (And, of course, similarly the other way around, plus DKIM covers some
> >> > cases when messages are relayed or forwarded.)  That speaks for "SPF
> >> > *or* DKIM".
> >> >
> >> > But "SPF *and* DKIM" -- requiring *both* to pass -- is technically
> >> > unnecessary at best, because of (1) above: DKIM should always pass
> >> > when SPF passes.  But where the harm comes is in cases of
> >> > mis-configuration, because now if *either* of them is misconfigured,
> >> > the whole thing fails -- neither of them serves as a backup for the
> >> > other; instead, the misconfiguration of either one causes
> >> > deliverability problems.  DMARC is damaged by requiring an
> >> > authentication situation that is unnecessary when things are properly
> >> > configured, and that is more fragile than what we've been using, more
> >> > susceptible to configuration errors than we've seen before.
> >> >
> >> > And I'm afraid that people will use it preferentially, *thinking* that
> >> > it provides better "security" -- surely, double authentication is
> >> > better than single, no?
> >> >
> >> > No.
> >> >
> >> > Barry
> >> >
> >> > On Tue, Jun 27, 2023 at 6:36 AM Alessandro Vesely <[e-mail]>
> wrote:
> >> >>
> >> >> On Mon 26/Jun/2023 20:13:53 +0200 Barry Leiba wrote:
> >> >>> I'm saying I don't want "and" to be an option, because I think it's
> >> >>> damaging to DMARC.  There is no reason anyone should ever want to
> say
> >> >>> that, and providing the option asks for misconfigurations because
> >> >>> people think it's somehow "more secure".  It's not more secure.  It
> >> >>> would be very bad for deliverability of legitimate mail and would
> >> >>> provide no additional security.  It would be a terrible mistake.
> >> >>
> >> >>
> >> >> I've been sporting spf-all for years, and seldom experienced
> bounces, mostly
> >> >> due to misconfigured secondary MXes.  Out of 39 domains whose posts
> to this
> >> >> list in the past year are still in my inbox, 14 have spf-all.  So,
> while I'm
> >> >> not the only one, not many published -all even though it may seem to
> be somehow
> >> >> more secure.
> >> >>
> >> >> I think it can be worth to compare SPF and DMARC.  Another sender
> policy a
> >> >> decade and an authentication method after.  What adoption, what hype.
> >> >>
> >> >> Both policies ask receivers to reject a domain identifier in some
> cases.  RFC
> >> >> 7208 explicitly suggests to consider whitelisting (Appendix D).
> DMARC provides
> >> >> for overrides but is less clear about how to handle exceptions.
> After SPF
> >> >> broke forwarding, the reaction was split between some changing
> identifier and
> >> >> turning to ~all; after DMARC broke mailing lists, between changing
> identifier
> >> >> and not altering messages.  In my limited experience, the ratio
> seems to be
> >> >> higher for DMARC than SPF, but I may be wrong.
> >> >>
> >> >> In theory, domains that currently have a strict DMARC policy and
> spf-all, 6 of
> >> >> the above, should have their messages blocked when either method
> fails, up to
> >> >> changing identifiers.  Why would it be so bad for deliverability to
> >> >> additionally require DMARC alignment, which is the difference
> between that and
> >> >> the "and"?
> >> >>
> >> >> And, it seems to me that an ESP not having a bloated SPF record
> could stop a
> >> >> good deal of DKIM replay by resorting to auth=dkim+spf.  Besides
> collateral
> >> >> deliverability problems, why wouldn't that work?
> >> >>
> >> >> Wht would "and" damage DMARC more than -all damaged SPF?
> >> >>
> >> >> I hope we can discuss detailed criticism rather than vague ostracism.
> >> >>
> >> >>
> >> >> Best
> >> >> Ale
> >> >> --
> >> >>
> >> >>
> >> >>
> >> >>
> >> >>
> >> >
> >> > _______________________________________________
> >> > dmarc mailing list
> >> > [e-mail]
> >> > https://www.ietf.org/mailman/listinfo/dmarc
> >
> >_______________________________________________
> >dmarc mailing list
> >[e-mail]
> >https://www.ietf.org/mailman/listinfo/dmarc
>
> _______________________________________________
> dmarc mailing list
> [e-mail]
> https://www.ietf.org/mailman/listinfo/dmarc
>