If you’re using mutt and you also have your private e-mail server running spamassassin, then you’ll see some spam messages still getting through the pipeline, with scores just below the default 5.0. So here’s the cure: use sa-learn on the server in order to train spamassasin bayesian filter and help it increase score on real spam.
You already have:
- offlineimap syncing your remote mailbox with a local folder from where mutt reads the messages.
- Junk folder in your IMAP mailbox, and server-side SIEVE scripts automatically move spam messages into this folder.
The big picture:
- Manually move the messages you want classified as spam into the Junk folder; offlineimap will automatically send them to the server upon next sync operation
- Have a cron-job on the server call the sa-learn utility on the Junk folder
Detailed steps:
file .muttrc
————
macro index ,j “:set confirmappend=no delete=yes auto_tag=yes\n=Junk\n:set confirmappend=yes delete=ask-yes\n” “Send message to Junk folder”
————
Log-in to the mail server then:
file /root/learn-spam.sh
———–
#!/bin/bash
sa-learn –spam [path to dovecot Maildir]/.Junk/cur
———–
#chmod +x /root/learn-spam.sh
#crontab -e
crontab file
———–
# run each day at 01:00 AM the learn-spam.sh script
0 1 * * * /root/learn-spam.sh
————
Voilà ! Each time you’re on a spam message in mutt, just hit ,j and the next day you’ll get more messages automatically classified as spam in the Junk folder.