Spam getting VERY low scores

Peter H. Lemieux mailscanner at replies.cyways.com
Mon Oct 2 15:42:23 UTC 2017


SA's Bayes engine put this meessage into the wrong category:

 > -5.00BAYES_05Bayes spam probability is 1 to 5%

so it appears your SA Bayesian database is screwed up.  Did you train 
the database with known spam and non-spam beforehand?  Did you use 
actual mail for the domain(s) you manage?

Here's a starting suggestion.  Turn of Bayes in SpamAssassin by setting

use_bayes = 0

in /etc/mail/spamassassin/mailscanner.cf.  If that helps with correctly 
classifying spam, then I'd delete the Bayes databases and rebuild them 
with new bodies of spam and non-spam messages.

Peter


On 10/02/2017 11:02 AM, Quintin S. Giesbrecht wrote:
> Here are the headers, spam report, and body of 1 of the emails in question:

> Spam report:
> -5.00BAYES_05Bayes spam probability is 1 to 5%
> 2.10FREEMAIL_FORGED_REPLYTOFreemail in Reply-To, but not From
> 0.25FREEMAIL_REPLYTO_END_DIGITReply-To freemail username ends in digit
> 0.50MISSING_MIDMissing Message-Id: header
> 0.01T_HK_NAME_MR_MRS
> 


More information about the MailScanner mailing list