[CLUE-Tech] Spam filters

[Randy Arabie]

On Tue, 1 Oct 2002, Dan Harris wrote:

> I actually prefer the keyword approach.  Since I switched to SpamAssassin, I
> have received ZERO spam in my inbox.  This is a reduction from over 30 a DAY..
> 
> The problem with the logfile analyzing types is that they never caught any
> spam because every message was from a different host.

My understanding of the statistical methods described in the link Dave posted
is that they are not based strictly on either keywords OR hosts.

The entire message is parsed and evaluated statistically on content and headers.

Here is a link to a thorough explanation:

	http://www.paulgraham.com/spam.html

> 
> Randy Arabie said:
> > On Tue, 1 Oct 2002, David Anselmi wrote:
> >
> >> This article
> >>
> >> http://catless.ncl.ac.uk/Risks/22.22.html#subj15
> >>
> >> references a spam tool based on statistics rather than keyword filters
> >> and block lists.  Might be interesting to those running spam filters.
> >
> > That is interesting.  I'm going to look into this more.  I read the "Plan
> > for Spam" doc recommended by ESR, which is his basis for the development of
> > bogofilter.
> >
> > A statistical based approach should be better than "scores" based on key
> > words.  The "Plan for Spam" went as far as suggesting each user build their
> > own baseline with a collection of their own non-spam email.
> >
> > However, I can't tell how far ESR's bogofilter goes, as the REAME and manual
> > are pretty light.  Guess I'll have to install and play.
> > --
> > Allons Rouler!
> >
> > Randy
> >
> > _______________________________________________
> > CLUE-Tech mailing list
> > CLUE-Tech at clue.denver.co.us
> > http://clue.denver.co.us/mailman/listinfo/clue-tech
> 
> 
> -Dan
> 
> 
> _______________________________________________
> CLUE-Tech mailing list
> CLUE-Tech at clue.denver.co.us
> http://clue.denver.co.us/mailman/listinfo/clue-tech
> 

-- 
Allons Rouler!
        
Randy