On Wednesday 25 June 2003 11:32, Per Björklund wrote:
> Wouldn't it be possible to expand all tests so that they can match all
> those p0rn, rem0ve and g1rls and so on, like the example I created below?
Listing every possible permutation could be quite difficult. How about
"normalizing" the text a
Per Björklund wrote:
Well, for example, I tried the attached porn spam (which uses a lot of 0's and 1's)
with 2.60-CVS,
but it still didn't score more than 0.6.
(I don't have razor, pyzor, bayes etc. on the machine I tried 2.60-CVS on so perhaps
that would have increased the score.)
This is what
I belive this is work for bayes. I just wondering why people turn it off ?
Why they try to use 'manual' bayes filtering ?
> > Wouldn't it be possible to expand all tests so that they can match all
> > those p0rn, rem0ve and g1rls and so on
>
> Perhaps uri references to real live p0rn sites, so tha
ONTRAST,HTML_MESSAGE,MIME_HTML_ONLY
autolearn=no version=2.60-cvs
/ Per
-Ursprungligt meddelande-
Från: Tony Earnshaw [mailto:[EMAIL PROTECTED]
Skickat: den 25 juni 2003 12:41
Till: Per Björklund
Kopia: [EMAIL PROTECTED]
Ämne: Re: [SAtalk] Target those 3l33+ typings
Per Björklun
Per Björklund wrote:
Wouldn't it be possible to expand all tests so that they can match all those p0rn, rem0ve and g1rls and so on
Perhaps uri references to real live p0rn sites, so that SA could watch
what was going on and learn for itself. Or, better, install a webcam in
your bedroom and wait
Wouldn't it be possible to expand all tests so that they can match all those p0rn,
rem0ve and g1rls and so on,
like the example I created below? Would it decrease performance to much? Or does it
have any other side effects?
It would at least make it harder for spammers to avoid spamassassin by o
