[lug] More Spam Please?

[Kevin Fenzi]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

>>>>> "The" == The Matt <thompsma at colorado.edu> writes:

The> <snip>
>> I do wonder about not using a spam collection though - what makes
>> you unique from the rest of us that you're afraid you'd misclassify
>> some incoming mail if SA was trained from it? Odd friends (well,
>> you do talk with us :) or are you afraid you'd miss the "Grow Your
>> Antenna To ENORMOUS lengths with our Herbal Concoction!  wx8347"
>> messages?

The> No.  The main reason is that I get quite a few large HTML
The> messages that SA *really wants* to make spam.  They are
The> daily/weekly newsletters and comics mails that have all the right
The> words to hammer SA.  The one thing I might do is make a copy of a
The> Spam collection and hand prune it.  The obvious spam is OK, it's
The> that stuff right at the edge that could kill the filter.  I'd
The> like to see how well the Bayes handles this.

In cases where the mail is coming from the same (known) email
addresses, you can add them to a whitelist for spamassassin. Ie, 

echo "whitelist_from newsletterplace at domain.com" >> ~/.spamassassin/user_prefs

Or stick that in the sitewide /etc/mail/spamassassin/local.cf. 

There is also a AWL feature (auto whitelist). If you tend to get non
spam from an address, then it gets a weighting to allow them to send
most anything. Likewise if you get spam from a specific address (ie,
support at microsoft.com) then they tend to get wighted more toward
spam. 

More info in the man page:

man Mail::SpamAssassin::Conf

The> Of course, the approx. 200 or so good HTML mails I'll send SA as
The> time goes on should teach it. :)

Yeah, that should also work... 

kevin

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.1 (GNU/Linux)
Comment: Processed by Mailcrypt 3.5.8 <http://mailcrypt.sourceforge.net/>

iD8DBQE+261/3imCezTjY0ERAjgnAJ4mvvvjylzhY6eg5u3dJeXWub0FtACeN48F
wNCaB3dlFThXQPSOeMvCQZc=
=pO++
-----END PGP SIGNATURE-----