spamassassin training

[IsaacT]

I have plesk 10.4.4 setup with spamassassin. Some of my users are getting a lot of spam, and we are attempting to train spamassassin better, however I'm not sure the best way to go about this.

From reading on here it seems that the best way to train spam is to move all emails that are spam into the spam folder in webmail. And then a cron job happens daily that will train spamassassin to consider these emails spam.

how much training does it take for spamassassin to learn? Meaning one of my users have moved hundreds of emails into the spam folder. and its been weeks, but the level of spam the user receives does not seem to diminish. They still receive ~50+ a day.

Also, does changing from qmail to postfix change anything with spamassassin? I didn't use to receive any spam, guess my training went well. However now that I changed to postfix from qmail because of the ipv6 issues i started receiving spam. only about 4 in the past 3 days. So its not like a floodgate opened for spam, was just wondering if there was anything linking this.

Lastly, is there any way to train spamassassin without having to use the webmail client, or imap? All my clients use pop email, so accessing the spam folder isn't possible. Do I just need to force them to use the webmail, or imap, or is there another way?

Thanks

 

[IsaacT]

another quick question about spamassassin. Does training work by putting emails in the spam folder even if your settings are to delete all spam messages? Or does having that option set not allow it to train since it wouldn't be using a spam folder?

 

[abdi]

For starters enable SPF and graylisting ...Do you have both of these enabled on your server?

 

[IsaacT]

Both of those are enabled. After some digging I think I found part of my answer. It looks like training via the spam folder only works if you have spam sent to that folder. If you have spam automatically deleted you can't train it more. That seems like a design flaw if you ask me.

 

[IsaacT]

Another question I have to this. Is there a way to reset training? so that it has nothing learned? Most of my spam is coming in as negative scores. I thought if I wiped out the traning and started fresh it might clean things up some.

also looking online it seems its possible to download a corpus of spam and ham files that can be trained into spamassassin. These files are from 2006, so while not current spam, they could help.

Does anyone know if resetting and trying to train all new would be worth it? or just keep moving things into spam with the hopes that eventually everything will get a high enough score?

If I can and should reset the training, how can I manually run the sa-learn function on the server. Since each email account does not have a user associated on the server, I wasn't sure how to specify the sa-learn for a particular email address. or if it just looks at the receiving email address when it "learns" the email message.

Thanks again for any information anyone can provide.

 

[barryc]

i have this same issue

anyone able to help?

 

[barryc]

sa-learn syntax

fyi, i found this:
http://www.digitalmurph.com/wordpress/?p=7

the commands seem to do something... but i'm not seeing my bayes_* files update their timestamps... so not sure if it's working or not yet. but just in case it helps.

i'll update in a few days if the spam stops...