Bayes and forwarded mail

[Joe Polk <li...@javelinux.com>]

If you setup a mailbox specifically for Bayes to learn and forwarded emails to
it, will Bayes sniff them out in a forwarded form? I assume you couldn't bulk
forward, but would this work otherwise?

--
<<JAV>>

Re: Bayes and forwarded mail

[Matt Kettler <mk...@evi-inc.com>]

Joe Polk wrote:

>If you setup a mailbox specifically for Bayes to learn and forwarded emails to
>it, will Bayes sniff them out in a forwarded form? I assume you couldn't bulk
>forward, but would this work otherwise?
>
No. If you feed sa-learn a forwarded mail, it will learn that forwarded
messages are spam.

Besides, the original headers can't be reconstructed from a normal
forward, thus proper learning is impossible. (And of course the body is
massively changed as well, with different mime encodings, etc). Although
a forwarded message may look similar to you, they look nothing like the
original to SpamAssassin.


The best option here is to use a little script to strip off attachments
and feed those to sa-learn, then have your users forward spam messages
as attachments. However, they must be able to do this with the full,
original, unmangled message with all headers, and all mime sections in
their original formats.