Michael Monnerie a écrit :
As there was the question of how to learn SPAM/HAM automatically, I
decided to release a mini script for testing:
http://zmi.at/x/spam_ham-checker
We use it like this:
- e-mail arriving is checked via SA
- detected SPAM is checked by human, and positives are put in a folder
named SPAM_yes, false positives to folder SPAM_no
- via cron we run above script:
# Find LDAP SPAM
0,30 * * * * root /spamscripts/spam_ham-checker
- this feeds SPAM/HAM to SA, learning it correctly. Mail is deleted from
folder after it's been learned.
Comments welcome.
if you delete after learning, make sure FPs are copied and not moved.
otherwise, you need to redeliver, but that would be a bit silly (just
because you sa-learn doesn't mean they won't be reclassified as spam...).
also, the permission issue must be handled correctly (in my case,
mailstore belongs to vmail, but SA is run as another user). I am still
looking for a way to not use root...
