How to train the spam filter?

How can you train the spam filter on Mail-in-a-Box? I don’t see any way of marking stuff as spam in Roundcube.

A while back Josh mentioned that by simply putting it in your Spam folder, it will start learning.

If you go to 'Special Folders" under “Preferences”, you’ll see an option to define which folder will be your spam folder.

Once you get an email, just move it over to your designated spam folder.

I prefer not to empty my Spam folder for a certain amount of time, in case another email from the same source comes in, the filter will know that there is one in Spam already, so this should go into spam as well. I do this largely because there are false positives here and there.

1 Like

You have to drag it in (or out) of the folder named “Spam”, but that may or may not be working.

1 Like

Howdy, I can’t find the original post on MIAB.

I had the same issue. I found on this forum a link to the Spamassassin Bayesian wiki and read all of it including the Kurt Yoder link at the bottom w/ his Perl script to report SPAM to the razor. This is less than ideal but does work. Over the last few months my SPAM box (Roundcube Junk directory) has slowed to a trickle (mostly the same SPAM). Inbox stays clean. Every morning I ssh into (box.) and run both scripts. The Kurt Yoder script I modified slightly to accept ctrl-C sigint Kill to stop the script from infinite loop (a few seconds to report all SPAM). A cursory inspection of my inbox to assure that it is clean then I run sa-learn script so Spamassassin will learn acceptable mail. So for Spamassassin to work requires both training and reporting. HTH.

1 Like

Pretty sure that’s not how it works. If the training works as described, the processed mails would become part of the bayesian corpus, regardless of whether you keep them in your spam-folder after they’ve been scanned, or not.

Yeah, that seems like a bit of a chore. I guess I’d rather deal with the little bit of spam that makes it through in the old-fashioned way :wink: