request for help

[KRFinch at dstsystems.com]

The best thing to do would probably be to have a special message "key" that
a designated list-person could forward the message with.  Special user gets
a piece of spam, and then forwards it onto the list with the "key".  Then
we script the filtration bits so that a message from user X with "key" Y
gets digested into the corpus of known spam.

We could open it up to everyone, but the problem with that would be that it
would probably skew the percentage numbers because two dozen people would
click on the same piece of spam and add it in.  It could also potentially
be abused because someone could effectively blackball someone else from the
list by just forwarding all of their messages on as spam.

No idea how we would implement all of that because I'm no email filtering
genius, but I know that it is almost certainly possible.

Another thought is that once we start having someone clicking through stuff
designating it as spam, I suppose it would be pretty academic to have those
known spam messages sanitized and forwarded onto other list members that
were putting together their own lexicons of spam.  People on the list that
were so inclined and technically capable could subscribe to the "rejects
list" and have those messages automatically added into their own respective
spam databases.  That way, one person reads it, and lots of people can
benefit.

- Kevin

                                                                                                    
                
                    Gerald Combs                                                                    
                
                    <gerald at ethereal.com>        To:     "Duston, Hal" 
<hdusto01 at sprintspectrum.com>                
                    Sent by:                     cc:     kclug at kclug.org                            
                
                    owner-kclug at marauder.i       Subject:     RE: request for help                  
                
                    lliana.net                                                                      
                
                                                                                                    
                
                                                                                                    
                
                    08/19/2002 04:26 PM                                                             
                
                                                                                                    
                
                                                                                                    
                

On Mon, 19 Aug 2002, Duston, Hal wrote:

> I have already been thinking about this after Kevin's posting
> as well.  It seems deceptively simple doesn't it?  I just may
> play around with something.

The Bayes' Rule stuff seems pretty straigtforward.  The user interface
might be a bit tricky, since most mailers don't have a "this is spam"
button.