[Mimedefang] using sa-learn in conjunction with mimedefang

Lucas Albers admin at cs.montana.edu
Tue Dec 23 20:35:17 EST 2003



Someone said...
> Am Di, den 23.12.2003 schrieb Douglas, Jason um 18:47:
> You can customize these values in your sa-mimedefang.cf, but a SPAM
> value of 2.7 is not likely to be auto-learned. It should be sure that
> the mail is really SPAM.
>
Their are defaults for which sa will not learn it as spam.
Don't learn at such a low score. You are setting yourself for a bad bayes
database.
I reccomend you learn at your reject level +3, or a minimum score of 8.
My values are:
Tag mail:5
Reject: 9.1
Learn with SA 12.0
Auto Report to dcc,pyzor,razor: 14.0
Relay-Test: 20.0

Just let the bayes database slowly but surely learn as it gets spam.
The sa developers spent a lot of time determining decent defaults, I would
reccomend you just follow the defaults.
You never want false positives on mail you learn from, as you have to
unlearn it.
It took 5 weeks for my system to get enough learned mail in it's corpus
(200 minimum) before it kicked in.



-- 
Luke Computer Science System Administrator
Security Administrator,College of Engineering
Montana State University-Bozeman,Montana



More information about the MIMEDefang mailing list