Hello David,

Looks like you have some duplication between your private rule set and
the SARE rules:

Monday, June 7, 2004, 5:32:42 PM, you wrote:

DBF> On Mon, 7 Jun 2004, Scott Rothgaber wrote:

DBF> Ran it here and it hit half a dozen tests, but they were all add-ins
DBF> (SARE, locally written, & Bayes) none of the stock distribution tests
DBF> hit. So this would be an argument to keep training your Bayes and
DBF> utilize new rule sources such as SARE and this list. ;)

DBF> Here's the rules that it hit here:

DBF> Content analysis details:   (20.5 points, 6.0 required, autolearn=spam)
DBF>  pts rule name              description
DBF> ---- ----------------------
DBF> ------------------------------------------
DBF>  2.2 SARE_HEAD_SPAM         Message headers used which identify spam
DBF>  5.5 L_BAD_HEADERS1         Headers that ony spam uses

If you check, you may find that L_BAD_HEADERS1 (based on that name) is an
older version of SARE_HEAD_SPAM, which included headers we found were
occasionally being used in ham.  SARE_HEAD_SPAM includes ONLY headers
that do not hit any ham in any of our corpora.

DBF>  5.4 BAYES_99               BODY: Bayesian spam probability is 99 to 100%
DBF>                             [score: 1.0000]
DBF>  2.5 SARE_RD_TO_BAD_TLD     URI: Redirect to bad TLD (info|cc|ws|biz)
DBF>  1.4 SARE_RD_YAHOO          URI: Uses unsecure Yahoo redirect
DBF>  3.5 L_URI_REDIR            URI redirector

L_URI_REDIR is probably (based on that name) is the predecessor to the
SARE_RD_ rule set, and has been superceded by the SARE_RD_* collection.

Remove L_BAD_HEADERS1 and L_URI_REDIR, and your installation may be a
little safer from false positives. That will drop this spam from 20.5 to
around 11.5, still an obvious spam.

Bob Menschel







-- 
Best regards,
 Robert                            mailto:[EMAIL PROTECTED]


Reply via email to