In
<198A730C2044DE4A96749D13E167AD375A2A05(_at_)MOU1WNEXMB04(_dot_)vcorp(_dot_)ad(_dot_)vrsn(_dot_)com>
"Hallam-Baker, Phillip" <pbaker(_at_)verisign(_dot_)com> writes:
Sure the bad guys will respond with a replay attack at some point,
but as has already been explained at some length it is a softball
attack that is not going to tax anti-spam schemes of any scale.
Uh, I don't think that it has already been shown that the replay
attack is going to be easy for anti-spam schemes to deal with.
If DKIM can't place strong controls on the sending of bulk email using
the signer's reputation, then I don't seem any value in it.
While some people say that spam can include things other than UBE, it
is really only UBE that causes the scaling problem.
If it is not bulk, then there is a limit to how many emails the
spammer can send. In the limiting case of everyone being spammers,
then everyone will receive as many spams, on average, as the average
person can compose.
If the message is solicited, then it doesn't make any difference if it
is bulk.
Yes, things like Razor/DCC/Pyzor can detect if a message is bulk, and
doing DKIM will make life *much* easier for them. Those systems don't,
however, do a good job of detecting if the message is unsolicited.
The principle challenge in establishing a free anti-replay scheme is
that generating the critical mass to redo DCC in an open fashion
without the censorship issues is not really going to be possible
until after the attack occurs. But by Web standards it is not a very
difficult or complex collaboration.
The other half of UBE detection, determining if the email was
solicited, is hard. I'm not sure, but it sounds like this is in part
what you mean by "censorship" since it is easy for people to falsely
declare a bulk email as being unsolicited, and thus giving the email a
bad reputation.
Right now, the "solicited" status is generally determined by things
like if the content looks hammy (bayesian, content filters, etc.),
whether the user has whitelisted the source, or whether the source has
a good reputation.
I thought the goal of DKIM (and the like) was to allow another,
better, way of determining the reputation of the source, and thus have
a good idea whether the email is solicited or not.
Again, if DKIM can't place strong controls on the sending of bulk
email using the signer's reputation, then I don't seem any value in
it.
-wayne