[extropy-chat] spam & spelling?

paul.bridger paul.bridger at paradise.net.nz
Sun Jul 25 22:21:01 UTC 2004

Good idea. To a certain, though, Bayesian filters currently do this.

Since no legitimate email would spell the names of drugs with numbers and 
punctuation embedded, these misspelled words will become indicators of spam 
(as you train the filter). Of course, spammers are very creative mis-spellers
and can also mess with the filter by delineating words differently (v-1-ag-ra 
..real..ch34p anyone?).

Brian Atkins wrote:
> I'm sure this has occurred to developers working on spam filters, and 
> perhaps it has even been implemented, but anyway:
> Why not judge email messages by spelling errors, and if they have too 
> many then they get marked as spam? This would seem to get around the 
> current problem of spammers coming up with all kinds of misspellings to 
> get around current spam filters. And most mail clients have a 
> user-extendable dictionary built in already for spell checking that 
> could be used.
> Now to catch a high percentage of spam you might have to set this kind 
> of filter to judge pretty harshly, and it might tend to catch legitimate 
> emails. So it will have to be used as only part of a larger spam filter 
> algorithm, but I do think it would help.
> Anyone know of any filters that use this technique?

More information about the extropy-chat mailing list