[cryptography] DKIM: Who cares?
iang at iang.org
Thu Oct 25 01:22:42 EDT 2012
On 25/10/12 14:34 PM, Peter Gutmann wrote:
> Zack Weinberg <zack.weinberg at sv.cmu.edu> writes:
>> Or perhaps the mere presence of a DKIM record is sufficient deterrent against
>> spam with forged From addresses at a particular domain, and that's the only
>> thing these organizations thought DKIM was good for.
> I think it's more likely that DKIM is affecting spammers so little (if at all)
> that they never really cared about it, and the organisations deploying it know
> that and don't bother doing anything more than going through the motions using
> the shortest (= lowest-overhead) keys. The thinking is that if DKIM had any
> effect on spam we'd have seen some sort of change in spam volume after it was
> deployed, but AFAIK there's been no effect on spam, just as SPF and who knows
> how many others have had no effect:
I think .. given that there is widespread confusion over the efficacy of
these systems, they've done us a favour. They've set the key at a low
enough level such that if the system were doing any good, it would be
attacked at that level. That would tell us something.
It hasn't been. It wasn't attacked by a bona fide attacker but someone
better classified as a researcher. Which tells us approximately .. nothing?
So what we are left with is a demo of how easy it is to attack, and how
uninteresting. In contrast, that mostly harmless system called SSL is
now being attacked on a few fronts, but is still set to be weak and
attackable on Android.
It is an oddity in the risk field that we can make keys so long such
they can't on paper be attacked. We can't do that with other
mitigations, crypto has this singularity when connecting to other
components, which makes the risk model wobble somewhat. This doesn't
mean we should do that, necessarily. Quite what it means, we may be
finding out now as various systems set at 512-1024 display their age.
> Having said that, if anyone at one of the DKIM-using organisations would like
> to contact me off-list to provide their point of view as to why toy keys were
> used, I'd love to hear about it. My guess it that it's a case of
> crypto-geeks : 0, operational considerations : 1, but there may be more to it
> than that.
> cryptography mailing list
> cryptography at randombit.net
More information about the cryptography