SpamAssassin at SMTP time with exim

[Danny Howard]

On Tue, May 07, 2002 at 03:28:51PM -0700, Joe Gross wrote:
> On Tue, May 07, 2002 at 02:48:35PM -0700, Danny Howard wrote:
> > 
> > I have a sneaky suspicion that a handful of spammers are already
> > checking their messages against SpamAssassin.  I have a feeling that
> > before long, SpamAssasin will lose its efficacy, thanks to an
> > ever-increasing spiral of spammers making SA-friendly e-mails, and
> > SA having to adapt to ever more ingenious heuristics.
> 
> I disagree, at least as long as SpamAssasin, or systems like it are
> used by the minority of users. For now SpamAssasin appears to work
> only with UNIX.
> 
> UNIX users are already less than 1% of spam recipients. UNIX users who
> use
[...]

Well, everyone on this list at least aspires to manage Large Unix
Installations, many of which are capable of supporting SpamAssassin for
large user communities, who don't themselves run Unix.  Therefor, the
strategies that we employ when implementing SA at an enterprise scale
are totally up for discussion.

Marc or Rick made the best response to my reservations in that SA relies
somewhat on collating information from different blackhole lists.  That
attribute of your message is difficult to munge, so we're less likely to
see spammers successfully competing with SAs heuristics.

I think the ultimate solution will be for anyone who receives unfiltered
spam to tag its characteristics in to a distributed network, which can
constantly tune the criteria used to determine a messages spamminess.
The first versions of this strategy already exist, I believe.  Vipul's
Razor?

-- 
http://dannyman.toldme.com/