Hi, I wrote this script some time ago. here is how it works:
I have an imap account that several users added to their email clients. whenever they get a spam message that didn't get identified - they move the message over to this spam account, and when they get a false positive, they move it to the ham account.
on the server, i run the script via cron and it learns from this account. the script doesn't have much comments, but it very easy to understand.just enter the configuration options in the begging and that should be enough.
its also pretty easy to modify to to work for any number of email accounts... and there are probably better ways to do it, but its done as it is and it works, so i don't see a need to redo it in any other way.
Hope it helps..btw, make sure that you get the bayes db one for all of the users, or else you will have to run this thing for each user separately... or maybe you would want that =)
Thiago Henrique wrote:
Hi, I have a MTA server with Postfix2.4.5 + Amavisd-New2.5.2 + SpamAssassin3.2.3. In another machine, I have a Mailbox server with Cyrus-Imapd2.3.11. The users are "virtual users" and are in a Ldapdatabase.We want to enable the learning Bayesian in SpamAssassin. I read enough doc of sa-learn but did not find anything about learning with messages remotely located. We can train spamassassin with messages in a folder remotely located? Thank you all in advance. Best Regards -- []'s Thiago Henrique Network Administration Digirati Networks K8 Networks
spamcollect.pl.gz
Description: GNU Zip compressed data
