Got this in my inbox this morning with a Spam Assassin score of 0.1: “If yöu Ã¥sk yöursëlf “Höw cÃ¥n it bë sö chëåp?”, thë Ã¥nswër is simplë: wë buy hugë quÃ¥ntitiës dirëctly fröm thë möst fÃ¥möust PhÃ¥rmÃ¥cy Pröducërs (whërë pëöplë cÃ¥nnöt buy) thÃ¥n wë chÃ¥rgë just önë cënt për dösë. This wÃ¥y yöu gët thë bëst pricës Ã¥ll övër thë wörld Ã¥nd wë sëll much mörë thÃ¥n öur cömpëtitörs.” I wonder how long it’ll take my bayesian filters to learn this trick. And no, they weren’t using CSS or a punk band.
It’d be good if you could have all characters that look like a certain character be converted to that character prior to application of the filters. “Ã¥” should become “a” and so forth. Seems silly to have to wait for the bayesian filters to catch up.
Bayesian filters and these spammers seem to be racing each other.
And the spammers are winning.
We non-anglophones have had to deal with this problem for some time. I get a huge amount of English spam for a small amount of English legit email, and a small amount of French/Spanish/German/Japanese … spam for a bigger number of legit emails in German and French (but not in other languages). Spamassassin deals with it alright, once I’ve trained it with the first samples of a new batch. But ultimately, Bayesian filtering is bound to be a problem, see this post by Edward W. Felten.
Heh… I find that hilarious. Who the hell writes all those nasty thingies above their letters? Haha! Silly people! :op
Reminds me of some MP…
Just entered my mind… I’m sorry. I’m just that way.