On 2010/03/11 6:32 AM, idahank wrote:
I use spamassasin 3.2.4 with amavisd-new 2.5 and sendmail 8.13.1 on Redhat Enterprise LINUX 4.8 Several of my network's users receive quarantine Emails from other providers that contain lots of spam-tokens without being spam. As different rules detect these mails as spam it is accidently learned as spam by autolearn. Whitelisting ist no option as it only puts a ham bonus of 100 points to the mail. Consequently spam tokens in these mails would be wrongly learned as ham by autolearn. A work around could be to detect these mail and let them be forgotten. Is there a way within spamassassin (or amavis) to exclude mails from a defined sender or subject from autolearn?
The bayes_ignore_from and bayes_ignore_to configuration options should do what you want. There's also the WhiteListSubject plugin in v310.pre if you'd prefer, but that can be abused.
For what it's worth, whitelisted addresses are excluded from Bayes autolearn.
grep -R noautolearn /var/lib/spamassassin/ |grep WHITE tflags USER_IN_SPF_WHITELIST userconf nice noautolearn tflags USER_IN_DKIM_WHITELIST nice noautolearn net userconf tflags SUBJECT_IN_WHITELIST userconf nice noautolearn tflags USER_IN_WHITELIST userconf nice noautolearn tflags USER_IN_DEF_WHITELIST userconf nice noautolearn tflags USER_IN_WHITELIST_TO userconf nice noautolearn -- /Jason
smime.p7s
Description: S/MIME Cryptographic Signature
