Koozali.org: home of the SME Server
Obsolete Releases => SME 7.x Contribs => Topic started by: alejandro on October 08, 2008, 12:14:48 AM
-
Hi all..
I would like to use sa-learn scripts to read spam/ham from an unique folder which all users will be suposed to send missclasified mail.
Actually It should be two unique folders for spam/ham for all users.
when any user forwards a mail to "spam@mydomain.com"or "ham@mydomain.com", files should be stored there and "learn as spam/ham" scripts could read and learn.
modifying sa-learn scripts is not the problem, but redirecting everything sent to "spam@mydomain.com" or "ham@mydomain.com" is where I have no clues.
Why I need this strange mechanism?
We don´t use imap folders nor webmail as mail client, and mails are deleted when downloaded to workstations as we don´t need/want to keep user files at server.
Spam filter is working pretty fine but we need to train and "finetune" spam filter with bayesian procedural.
I know it could result in something out of standard use but maybe a usefull add-on to spamassassin aswell.
Many thanks in advance for anyone tht could give me a clue on this
Ale.
-
Hi all..
I would like to use sa-learn scripts to read spam/ham from an unique folder which all users will be suposed to send missclasified mail.
Actually It should be two unique folders for spam/ham for all users.
when any user forwards a mail to "spam@mydomain.com"or "ham@mydomain.com", files should be stored there and "learn as spam/ham" scripts could read and learn.
modifying sa-learn scripts is not the problem, but redirecting everything sent to "spam@mydomain.com" or "ham@mydomain.com" is where I have no clues.
Why I need this strange mechanism?
We don´t use imap folders nor webmail as mail client, and mails are deleted when downloaded to workstations as we don´t need/want to keep user files at server.
Spam filter is working pretty fine but we need to train and "finetune" spam filter with bayesian procedural.
I know it could result in something out of standard use but maybe a usefull add-on to spamassassin aswell.
Many thanks in advance for anyone tht could give me a clue on this
Ale.
I think you might have an issue here by forwarding it to a new address as mail headers will not be the original mail headers anymore, which makes SPAM/HAM training give false positives I think.
-
One possible workaround:
1. set your POP client to leave mails on the server 5 days.
2. If a user gets a batch of spam that they want learned, they can log into webmail within 5 days and move it to the LearnAsSpam folder.
-----------------------------------------------------
For HAM that is marked as spam, use this:
http://wiki.contribs.org/Sme-unjunkmgr