Hi Joe,
jbi joe wrote:
IM getting way too much spam. I have tried to train the bayesian analysis
mailet by feeding it spam and notspam. However, it does nothing.
I haven't looked too closely but I did spot that your config file has
the following setting: -
<ignoreLocalSender>true</ignoreLocalSender>
This setting will mean that any email sent to your server in which the
sender's address is one of the domains that your server serves will be
considered local and will completely bypass your Bayesian filter.
Unfortunately this is all too easy to forge and a lot of spam simply
uses the victim's email address as the sender address.
Simply change this value to false and restart James.
You may also like to get your hands a little dirtier and login to your
database server to see if the ham and spam tables have anything in
them. If not then maybe your ham or spam messages aren't being added to
the corpus for some reason.
Another quick check is to look at the emails and check that the
X-MessageIsSpamProbability header is being set. If the value is missing
that could also indicate that messages aren't passing through the filter
for some reason.
Regards,
David Legg
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
