Question about how bayes learns from HTML messages

[David Hooton <dj...@platformhosting.com>]

Hi All,

 

I was just wondering how bayes inspects html messages?  Does it look at the
text or does it learn from the code?

 

The reason I'm asking is that we regularly get defanged html messages which
present as a blank page.  I'm concerned that training these as spam may not
be a good idea if it doesn't inspect code rather than text.

 

Regards,

 

David Hooton


========================================================================
   This message has been scanned for spam & viruses by Mail Sleuth.
   To report SPAM forward the message to:    spam@mailsleuth.com.au
   Mail Sleuth                                www.mailsleuth.com.au
========================================================================

Re: Question about how bayes learns from HTML messages

[Matt Kettler <mk...@comcast.net>]

At 11:22 PM 4/1/04 +1000, David Hooton wrote:

>Hi All,
>
>
>
>I was just wondering how bayes inspects html messages?  Does it look at 
>the text or does it learn from the code?

IIRC, Bayes learns from the same body that "body" rules are applied to. 
This body has had base64 decoding, QP decoding, all HTML tags removed, and 
end-of-lines removed.

Thus, SA's Bayes learns from the text, not the code. And that's a good 
thing, as it makes the spammer obfuscation technique of stuffing HTML tags 
into the middle of words completely ineffective against SA's bayes engine.