Michael W Cocke wrote:
I've got every ruleset & blacklist available and I'm still getting
buried - the bayes poison in all of the recent spam has wrecked that.
Does anyone see a reason why I can't assume messages with blank
subjects are junk?
(counter) examples are available on this list (see a message sent on
2006/07/27) and on other lists. I've also seen many corportae mail with
empty subject (forgotten, or considered irrelevant by the sender). It
even happens to me from time to time (delete the subject to replace it,
then see an error in the body, switch to correct the body, then forget
that the subject was deleted....).
you'd better look for other patterns. If your bayes isn't performing
well, trash it and retrain SA using manually inspected mail. This should
cut a lot of spam.
Also, if you receive legitimate mail that has spam patterns
(really-opt-in newsletters), you'd better create special addresses to
subscribe these, and not use these addresses for bayes training.
Also, I've got an idea about maybe doing an
nslookup on the envelope sender domain and junking anything without an
entry. I'm probably missing something that I should consider,
especially on that last one. Would anyone care to educate me what I'm
missing?
you can reject senders if domain doesn't exist for sure. do this in your
MTA. your MTA should return a temp failure in case of dns temp failures
though. and you'd better get your DNS setup correctly working (have a
cache dns on your mail server or on another box with good connectivity.
