From: <[EMAIL PROTECTED]>
Sent: Monday, 2008, August 18 11:08
On Mon, Aug 18, 2008 at 1:59 PM, John Hardin <[EMAIL PROTECTED]> wrote:
On Mon, 18 Aug 2008, [EMAIL PROTECTED] wrote:
Long time SA user here. I have googled much for an answer for this. I
have
a few email addresses that are clearly now spam only. I would like to
blacklist them and use them as a honeypot to help train my Bayes through
autolearn, does anyone have any suggestions on how to do this?
Training from those users' mailboxes is pretty straightforward using
sa-learn in a script run from cron. I wouldn't worry about trying to get
autolearn involved.
--
I guess I should have been more clear. Most of these email addresses have
been forwarding into a junkemail gmail account and have no 'local'
mailboxes
of their own, and I would like to keep it that way. I am also using
MailScanner if thats of any use. Basically I want (yes I dont care if a
spambot gets this addy) [EMAIL PROTECTED] to always only be recognized
as
spam content and have it learned. Since there isnt a mailbox how would
this
best be accomplished?
I believe you could "blacklist_from". That would train SpamAssassin's
Bayes filter - if you use global (shudder) Bayes. As for automating
building a block list you'll have to pester someone else into providing
a clue.
If you use per-user Bayes then you have more of a problem because you
need to train it from messages not sent to the user. For that you will
need a local spam repository that operates as a fifo buffer. How you
implement that is an exercise for the student.
{^_^}
