Hey everyone,

        We're currently coping with a false-positive crisis that's
sweeping our email with 2.60, mostly due to scores of the Bayes filter.
We run SA site-wide on an incoming MX host, so individual users do not
have access to train the Bayes database.  Moreover, our primary client
program is Pegasus Mail for DOS, which provides no real way to get raw
messages out unmodified (it hoses CR/LF, forces line wraps, and cat's
MIME parts together).

        So I'm going through some of our Bayes tokens trying to decide
if I should dump the current database and start over.  I've noticed
really bad things like this:

0.892   381     112     1069183901      HTo:[EMAIL PROTECTED]
0.905   75      19      1069183901      HTo:[EMAIL PROTECTED]
0.997   17      0       1069183901      HTo:[EMAIL PROTECTED]

This looks really horrible!  Just by virtue of my boss's email having a
"To: [EMAIL PROTECTED]", it'll almost certainly be tagged as spam.  The
database is trained with nham=13685 and nspam=5652.  Autolearning is
enabled and has default threshholds.

This is alarming at first.  But when I think about it, and I realize
that most of us get more spam than ham - Bayes is right.  Unfortuantely,
that's really, really the wrong thing to do.  Is there a way to excempt
some headers from processing?


-- 
Ross Vandegrift
[EMAIL PROTECTED]

A Pope has a Water Cannon.                               It is a Water Cannon.
He fires Holy-Water from it.                        It is a Holy-Water Cannon.
He Blesses it.                                 It is a Holy Holy-Water Cannon.
He Blesses the Hell out of it.          It is a Wholly Holy Holy-Water Cannon.
He has it pierced.                It is a Holey Wholly Holy Holy-Water Cannon.
He makes it official.       It is a Canon Holey Wholly Holy Holy-Water Cannon.
Batman and Robin arrive.                                       He shoots them.


-------------------------------------------------------
The SF.Net email is sponsored by EclipseCon 2004
Premiere Conference on Open Tools Development and Integration
See the breadth of Eclipse activity. February 3-5 in Anaheim, CA.
http://www.eclipsecon.org/osdn
_______________________________________________
Spamassassin-talk mailing list
[EMAIL PROTECTED]
https://lists.sourceforge.net/lists/listinfo/spamassassin-talk

Reply via email to