Hello!
I'd like to combine the Bayesian filter with TMDA server-wide. I
currently have the spamprobe spawned by procmail (acting as the local
delivery agent) on each message using a server-wide database.
This works Ok, but there are false-positives every once in a while. I'd
like to give the messages flagged as spam by spamprobe a chance through
the challenge-response mechanism, such as that of TMDA.
Here is my current rule, which calls spamprobe for all reasonably sized
messages (spam is rarely bigger than this):
:0
* < 500000
{
SCORE=`/opt/bin/spamprobe -8 -d /some/path/.spamprobe score`
:0 fhW
| formail -I "X-SpamProbe: $SCORE"
:0:
*^X-SpamProbe: SPAM
/some/spam-folder
}
Instead of the /some/spam-folder sub-rule, I'd like to run tmda-filter
(?), that will check the sender against the black list, and generate the
challenge for the potentially legitimate senders to respond to.
I also need a rule, that will run the carantined message through
'spamrpobe good' (to adjust the Bayesian database) should the successful
response come back.
Has anyone done this already? Can you share the procmail rules for this?
I don't think, I want to use the white list, because the current virii and
spam can pretend to come from a legitimate user...
Thanks!
-mi
_____________________________________________
tmda-users mailing list ([EMAIL PROTECTED])
http://tmda.net/lists/listinfo/tmda-users
