Spamassassin settings for bayesian filtering

networking November 12th, 2006

SPAM fighting (A talking to myself, want to get this in google post…)
I’ve been getting a HUGE number of SPAMs slipping through my SPAMASSASSIN filter, and have been trying to figure out how to combat it. SPAMs are getting flagged as BAYES_00, which gives them a negative score.

I was moving them into the SPAM folder, then running sa-learn nightly to train.
I’ve started moving those SPAMs into a separate folder, then running sa-learn—forget to forget the scoring, then moving them back into the SPAM folder and re-running sa-learn.

I’ve also tried changing the score for BAYES_00 from negative to 0 by adding the following line to my user_prefs file:

score BAYES_00 0.1

I should look at getting my addresses out of Outlook and into a whitelist format for Spamassassin.
So, the process looks like:

Move mis-filed HAM from SPAM to SPAM-moved

/usr/bin/sa-learn—forget—dir /home/kweiske/Maildir/.SPAM-moved/cur
/usr/bin/sa-learn—spam—dir /home/kweiske/Maildir/.SPAM/cur

Move the mis-files HAM from SPAM-moved to Inbox

/usr/bin/sa-learn—ham—dir /home/kweiske/Maildir/cur

Fingers are crossed.

There’s a good online resource regarding Bayesian filtering at, a free chapter in an online book.

Leave a Reply

You must be logged in to post a comment.