At 08:48 AM 7/31/2005, Loren wrote:
My guess, without looking at the rules in question, is simply that a smarter spammer played around until he found two specific mis-spellings that would not be caught by the obfuscated drugs and body parts tests, and then used those and only those two.
Exactly.. The existing DRUGS_ERECTILE rules don't look for "viagr" as a possible mis-spelling. There's another common variant, viigra, that gets missed too. If I ever have any spare time again, I'll look at tweaking the rules. If you check the archives I posted a modified __DRUGS_ERECTILE1 to catch the double-i variant.
Of course, all this said, there are relatively few variants missed by the drug rules, and in the interim a little bayes training should go a long way. It's not like these words will ever appear in ham so they should be quite learnable.
