[Mimedefang] Bayesian, please help

Lucas Albers admin at cs.montana.edu
Tue Dec 9 12:04:11 EST 2003


I will keep stating this to everyone who posts this sort of information.
Don't just drop email, do an action_bounce and and action_drop if it is a
detected virus.
This is the PROPER thing to do.
I also do an action_quarantine on everything, then I delete emails that
are more then a week old.
This adds some minor disk usage to your system but will save you work if
someone wants to recover something.

When you first implement sa start with high numbers then work your way
down, this allows you to learn what normal sa email scores are.
Bayesian is automatic, you just need to enable it in sa-mimedefang.cf
Don't pimp it up just to get it working, just let it learn on it's own.
Go read about in the sa website.
If your bayes learn scores are too low you will get crap in your bayes
database.
Look on the archive for how to enable bayes.

--luke

> Got MIMEDefang up and running, 2.39 with SpamAssassin 1.6.  Drops emails
> with a score over 8, marks anything between 4 and 8 as Spam but still
> delivers it.  I'm still getting some good emails marked as spam though
> so I'd like to use the Bayesian filter to "learn" what is spam and what
> is ham.



More information about the MIMEDefang mailing list