Hello Rob,

Friday, November 19, 2004, 6:40:41 PM, you wrote:

RB> I am currently using SARE_OEM SARE_GENLSUBJ SARE_GENLSUBJ_ENG SARE_HTML1
RB> SARE_HTML2 SARE_HEADER1 SARE_HEADER2 SARE_HTML_ENG SARE_BML SARE_FRAUD
RB> SARE_SPOOF SARE_UNSUB SARE_RANDOM SARE_TOP_200 and BOGUSVIRUS as my 
rulesets.

Are you running SARE_HTML1 and SARE_HTML2 without SARE_HTML0?
SARE_HTML0 is the more powerful rules file in that family, of which
SARE_HTML1 is the less effective smaller brother.

RB> All I want to do is push the scores into the spam range. And frankly I 
think I
RB> could lower the bar, too. Are their rulesets that might help, or custom 
rules
RB> that I could write, and as a single user I don't need perfection, I just 
want
RB> something like a 95% catch ratio instead of the 60% I am currently getting.

Check into the SARE_HEADER family also, and SARE_SPECIFIC. But even
without these, 60% seems awfully low.

RB> X-Spam-Status: No, score=3.1 required=5.0 tests=ALL_NATURAL,BAYES_99,
RB>     HTML_IMAGE_RATIO_04,HTML_MESSAGE autolearn=no version=3.0.1

If your Bayes database is well trained, bump up the score for
BAYES_99. I run with Bayes_99 = my required-hits threshold.

I also notice you don't have any of the SURBL or other Network tests
showing. If you can enable network testing I suspect you'll catch a
lot more of your spam.

Bob Menschel


Reply via email to