[ietf-dkim] Re: DKIM signature can mean it's safe to generate bounce?

Douglas Otis dotis at mail-abuse.org
Wed Sep 26 15:57:23 PDT 2007


On Sep 22, 2007, at 5:56 PM, Frank Ellermann wrote:

> Scott Kitterman wrote ten weeks ago:
>
>> On Friday 06 July 2007 20:09, Dave Crocker wrote:
>>> It seems to me that if a message has a DKIM signature and the  
>>> signing domain matches the domain in the rfc2821.MailFrom  
>>> command, then it is safe to generate a bounce message to that  
>>> address.
>
>>> By 'safe' I mean that one can be confident that the mail will not  
>>> go to an unwitting victim of a spoofed address.
> [...]
>> I expect if limited to the case where 2821 Mail From domain is the  
>> same as the signing domain it would likely be reasonably effective.
>
>> SPF Pass would (if available) give you the same or better confidence.
>
> Better because you'd already have this confidence after SMTP MAIL  
> FROM without DKIM crypto looking at the 2822 header.  SPF also  
> allows to aggregate more than one "sending provider", that's rather  
> tricky with DKIM or BATV.

This can be done with DKIM using a strategy that can bridge policy  
domains.

See:
http://tools.ietf.org/wg/dkim/draft-otis-dkim-tpa-ssp-01.txt

Unlike SPF, the tpa-ssp method scales without invoking hundreds of  
DNS transactions.  This draft extends the current SSP draft.

> OTOH you'd never see an SPF PASS behind "taditional" (aka "broken  
> by 1123 5.3.6a") forwarding, where Dave's approach based on DKIM  
> would still work.  The scenarios aren't equivalent, Dave's method  
> is more or less limited to cases starting out with 2822-From = MAIL  
> FROM, but after that it survives forwarding.
>
> If a mailing list manages to invalidate the DKIM signature Dave's  
> approach won't work.  But the list could very easily guarantee it's  
> own SPF PASS, so in practice receivers can handle most sound cases.

The use of the From is sure to create problems for mailing-lists.  As  
with SPF, SSP also lacks policy scope which is unfortunate.  This  
could be repaired by extending SSP policy with drafts like TPA.   
Unfortunately, these extensions are unlikely to be incorporated and  
more convoluted work-arounds will likely be used.

-Doug


More information about the ietf-dkim mailing list