Sipos Gabor schrieb:
Hi
i am using spamassassin as mail-filter, it tags me about the 40-50% of the spam.
I've use sa-learn on spam and ham messages, (i've learned about 3000
^^^^^^
spam messages)
I use 5 score to tag a spam but most of spam not detected is
recognized only by HTML_MESSAGE and reports 0 scores....
You could adjust the score for HTML_MESSAGE (or adjust it for other
rules that "hit" your spam) ...
(afaik default is 0, in later versions too ...)
*******************
X-Spam-Checker-Version: SpamAssassin 3.1.0 (2005-09-13) on unixmail.xxxx
X-Spam-Level:
X-Spam-Status: No, score=0.8 required=5.0 tests=HTML_MESSAGE,INFO_TLD
autolearn=no version=3.1.0
*******************
3.1.0 is "outdated" perhaps you could use a more recent version?.
sa-update?
What do you do with this kind of spam?
bayes will not work until you have trained it with at least 200 NON-spam
messages.
perhaps he has trained it on ham too, as stated above?
"spamassassin --debug"
or "amavisd-new debug-sa" (if used with amavis)
and a search for "bayes" would help ...
Gabor Sipos
hth
MH
