SA 3.0.1 and Bayes poisoning issue?

Matt Kettler mkettler at EVI-INC.COM
Fri Nov 5 20:42:31 GMT 2004

At 11:21 AM 11/5/2004, Jeff A. Earickson wrote:
>I have in my spam.assassin.prefs.conf file the following:
>#---tweaks to combat neg bayes poisoning
>score BAYES_00 0 0 -0.05 -0.05
>score BAYES_01 0 0 -0.04 -0.04
>score BAYES_10 0 0 -0.03 -0.03
>score BAYES_20 0 0 -0.02 -0.02
>score BAYES_30 0 0 -0.01 -0.01
>left over from the SA 2.64 era.  I'm wondering if I should remove
>this now that I'm running SA 3.0.1.  Comments anybody?

If you had bayes poisoning problems serious enough for you to have to do
that, upgrading spamassassin will definitely NOT fix the problem.

If you've got more than 0.01% of your spam matching BAYES_00 that's a
serious mis-training problem, and there's nothing a SA upgrade can do to
fix that. AFAIK no version of SA has ever had bayes poisoning problems if
properly trained, and mine certainly never has (I've been using bayes since
2.53 came out)

Bad training is bad training. Code can't fix bad training. Wipe out your
database and start over.

And beware.. Never rely on autolearning alone. You've got a good chance of
the bayes DB starting off on the wrong foot and getting stuck there. At the
very least hand-seed a handful of emails into sa-learn when you start off.

------------------------ MailScanner list ------------------------
To unsubscribe, email jiscmail at with the words:
'leave mailscanner' in the body of the email.
Before posting, read the MAQ ( and
the archives (

More information about the MailScanner mailing list