Andy Blanchard wrote:
Does anyone have a suggestion how to raise scores on messages like this the next time? I trained Bayes with it, of course and integrated the alternate spelling in my V... rule, but besides that I'm at a loss right now.
Looks like the spammer's solution to the detection of their "insert random punctation" approach is to insert random letters instead. It's highly likely that this approach is going to generate a *lot* of uncommon letter sequences, so I'd look to trigger on that. IIRC someone has already produced set of regular expressions based rules for letter sequences that do not occur in English, but I can't locate it on Google. Anyone got a link, because I think it may be time to add it to the collection...
Andy
The only person to have all his work done by Friday was Robinson Crusoe
In the meantime I integrated the Tripwire rules pointed out by Thomas Kinghorn and others. Now I'll wait and see... Maybe you should do the same?
Regards Ralf G.
