[CentOS] Trying to control the torrent of spam...

Scot L. Harris webid at cfl.rr.com
Sun Aug 28 20:32:09 UTC 2005


On Sun, 2005-08-28 at 12:11, Preston Crawford wrote:
> On Sun, 2005-08-28 at 11:21 -0400, Scot L. Harris wrote:
> > On Sun, 2005-08-28 at 11:01, Preston Crawford wrote:
> > 
> > > 
> > > An IDEAL setup would be one where spam gets auto-forwarded to the FTC's
> > > spam email address and to BlueSecurity.com. But I don't know if such a
> > > program exists for Linux.
> > 
> > Auto forwarding of such messages is not a good idea.  If there are any
> > false positives you could end up having legit email marked as spam and
> > sent to these third party systems.  You should have some kind of
> > confirmation before sending on such email.
> 
> That's a good point. That's just me speaking in frustration. When the
> spam turns to phone calls on your CELL PHONE it starts to get really
> maddening. It makes you want to change your phone number and your email
> address every 6 months. Which, if you're a software developer and a
> contractor (like me), isn't very practical. People have to know how to
> reach you.
> 
> > Depending on the type of spam you are getting you want to make sure
> > spamassassin bayesian filters are working.  I think you need to have fed
> > it about 200 spam messages before it kicks in.  You may also get good
> > results using some of the alternative rule sets provided by
> > http://www.rulesemporium.com/.  
> 
> I think they are, but maybe I'm not marking enough stuff junk in
> Evolution (which I assume uses spamassassin).
> 

That is my understanding as well.  I have been running spamassassin as a
filter in an older version of evolution for the past couple of years. 
The bayesian filter catches all but a handful of spam that is sent to my
accounts.  You will need to mark enough items as junk (about 200 I
think) to get the bayesian filter to start tagging spam.  The big
benefit of this is that it learns what you think is spam not someone
else's idea.  It just takes enough samples to make it work as expected.

> > If you were running your own MTA I would recommend you implement
> > greylisting.  This would eliminate >98% of spam with little or nor
> > impact to your MTA servers resources.  
> 
> I'm using fetchmail + pine and then evolution. Sendmail is somewhere in
> there as it comes with CentOS. Not sure where to implement the
> greylisting, though. I'll have to look that up.

As I expected, in your setup you won't be able to setup greylisting. 
Your ISP would have to set it up on their MTA.  Your best option is to
get bayes trained and possibly add a few selected rule sets to
spamassassin from the rules emporium as mentioned before.




More information about the CentOS mailing list