Anti-spam methods (was: Spam-Filter-Free Options)

[Jarod Wilson]

On Wed, 2008-05-07 at 12:03 -0400, Steven W. Orr wrote:
> On Tuesday, May 6th 2008 at 15:20 -0000, quoth Ben Scott:
> 
> =>On Tue, May 6, 2008 at 3:07 PM, Bill McGonigle <bill at bfccomputing.com> wrote:
> =>> ... DKIM ...
> =>
> =>  DKIM won't stop spam.  DKIM may make it easier to identify mail
> =>which came from senders you want to receive mail from.  There's a
> =>difference.  DKIM won't help if you want to receive unsolicited mail,
> =>or if you want to receive mail from systems which also tend to be spam
> =>sources (like the big web mail providers).
> =>
> =>  Anyone here using Maia Mailguard (http://www.maiamailguard.com/)?
> =>It purports to be a front-end mail filter with user-friendly web GUI
> =>and per-user capabilities.  I haven't had a chance to try it out
> =>though, despite wanting to for about two years now.  If it works, I'd
> =>love to stick it in front of our Exchange server at work.
> 
> I think I'm running one of the best solutions available: sendmail + 
> spamass-milter + spamassassin.
> 
> My sendmail config has a few tweaks to pick off the low hanging fruit.
> Then spamassassin does a great job at getting the rest.
> Spamass-milter allows me to run spamassassin on the message while it's 
> incoming and to reject it before reception even completes. That way the 
> badguy knows I rejected him.

While its ideal to be able to reject the mail prior to placing it in
your received queue and letting the spammer know its been rejected, it
doesn't scale particularly well on high-volume mail servers.

What I really like (and hate at the same time...) about maia is that it
provides a user-friendly web-based frontend to examine your quarantined
spam and rescue anything falsely labeled (which also whitelists the
sender and gets used to train the bayesian filter), and also to examine
mail that was labeled as non-spam, and let maia know what it missed
(again training the bayesian filter). Its a bit of overhead to look at
it, though you're not required to (caches expire automatically after
configured intervals).


-- 
Jarod Wilson
jarod at wilsonet.com