theo wrote:
> Is there an algorithm to find out faked email addresses in the form:
>
> WxOHT@<host>
> rokqawmwrp@<host>
> jhmtnr@<host>
> xwldsxu@<host>
>
> By reading these I'm 99% sure that these are fake addresses and that
> e.g. "rokqawmwrp" is not a name for anything or anybody in any
> language.
>
> Is there a way (algorithm) to do this programmatically? Getting a
> probability would be enough...
>
> Thanks
> Theo
interesting thought :)
Ok, let's wildguess a bit <g>
you could create a heuristical function to give you the probability
that a direction is a spammer based in a rule that indicates the
approximate distribution between vowels and consonants in a "normal"
word.
Another rule of thumb could be that it is very unlikely to have in a
word 4 consonants, a vowel, 3 consonants and so on.
Surely this function would need to take in consideration that not all
languages have the same distribution, although inside a group, like the
germanic or the latin languages, it is more or less equal.
Combined with white lists of directions and hosts (to allow all mails
coming from thatcompany.com to pass) and an algorithm that decides
based on that probability it could perhaps work.
Still, you have to consider those cases when the real e-mail addresses
really appear to be spammers but are not.
For example, I know that in Austria some people have as surname
Gschwandtner (creepy isn't it?), which surely such an algorithm would
consider as the king of spammers <bg>
Just my 2 cents ;)
--
Best regards :)
Guillem Vicens Meier
Dep. Informatica Green Service S.A.
www.clubgreenoasis.com
