Training SpamAssassin

Spamassassin has the ability to "learn" what is spam along with what is not spam with the sa-learn command.
http://spamassassin.apache.org/full/2.6x/dist/doc/sa-learn.html

Simple example.

• Make sure you have spamassassin running for you and some spam is being trapped. Learn How Here
• Create a mailbox folder with your IMAP client called salearn.
  (creates a file in /user/<username>/salearn or /user/<username>/Mail/salearn depending on your IMAP client setup. Make sure you know this for a later step)
• Move undetected spam from your inbox to salearn. This will build up.
• Depending on your volume, the next step could be every few days at first or a few weeks.
• SSH to arctic and run the command: sa-learn --spam --mbox /user/<username>/Mail/salearn
• This tweaks the spam filters to learn from the file salearn.
• If you use mh mail (each email is a number under a directory), then ignore the --mbox argument.
• The results will output how many emails were examined and the number of messages it learned from.
• You can clear out the salearn mailbox and start over.

Use this simple example to tweek the spam filter. For better results, teach it non-spam.

• Be sure to clear out all spam from your inbox.
• Be sure to know the location of your inbox (/var/mail/<username>, /user/<username>/mbox, other depending on your unique setup)
• "ls -l" those locations to find the latest time stamp, the directory with the most recent timestamp is your inbox.
• SSH to arctic and run the command: sa-learn --ham --mbox /var/mail/<username>
  (notice: --ham instead of --spam)
• Output will be similar to previous example.