Today the only spam which slipped thru was a short one with some URLs in HTML and then this at the end (see below):I opted for resetting the bayes database every couple of weeks - this kind of thing tends to make bayes more and more deranged... unfortunately it seems to be a weakness of the bayes method.
I think this is not only trying to trick Bayes in thinking it's ham but also tries to spoil the Bayes ham recognition if I teach that message as spam to sa-learn because Bayes will now add some spam percentage to BEGIN PGP SIGNATURE or "rocky" or "comparing" etc.
Or am I wrong?
What should I do? Teach the complete message to Bayes as spam or remove this stuff before teaching it?
Description: S/MIME Cryptographic Signature