On 8/4/2014 2:18 PM, RobertGrimes wrote:
I a bit new to all this so please excuse my lack of knowledge -
I am running hMailServer which is using spamassissan. I am getting some
emails through that I would think should be spam but show a very low score
with my setup. I would like to try to run the exact same email through
spamassassin multiple times with adjustments in parameters until I can get a
bit higher hit rate. If I take the message after it has gone through
hMailServer via SA, there are extra headers that SA has put in even if it
got a very low score like 0. If I try to run that message through SA via a
command line, it is now tainted with the new SA headers so I am not sure the
validity of the score now. The headers that were added, even though
hMailServer didn't ask for them look like:
X-Spam-Checker-Version: SpamAssassin 3.4.0 (2014-02-07) on server
X-Spam-Level: ***
X-Spam-Status: No, score=3.2 required=5.0 tests=BAYES_00,HTML_MESSAGE,
RCVD_IN_BRBL_LASTEXT,RCVD_IN_PSBL,RDNS_DYNAMIC autolearn=no
autolearn_force=no version=3.4.0
A better question is, what is a good methodolgy for tweaking SA parameters
and knowing you are actually starting with the same message and getting a
new score?
Any ideas would be appreciated. Thanks in advance.
X-Spam headers can be ignored. SA will automatically strip them before
processing the message.
The biggest problem I see with the sample message is the BAYES_00
score. This indicates that the Bayes scoring engine is confident that
this message is NOT spam. When this appears in obvious spam messages,
it indicates a training problem with your Bayes DB -- Spam messages have
been trained as ham (non-spam). Double-check your procedures for Bayes
training. Are you doing any manual training? If so, make sure that you
are training the right messages as ham and spam. Also, make sure you
are running sa-learn as the same user that SA is using.
--
Bowie
