[Spambayes] my pet theory on spam and language got proven today

Seth Goodman sethg at goodmanassociates.com
Mon Oct 30 01:36:27 CET 2006


skip at pobox.com wrote on Sunday, October 29, 2006 3:25 PM -0500:

>     Colin> I've tried this in Firefox with the filter "image/gif". 
>     So far, Colin> it seems to work.
> 
> The canonical example of where this won't work is a parent sending
> baby pix to Grandma and Grandpa.  As an individual you might decide
> that throwing out all mail containing images (or at least GIF
> images) is worth the risk of lost mail, but it's far from a general
> solution to image-based spam.

That depends on how it is implemented.  For example, Outlook 2003 is
supposed to render images in messages from correspondents listed in
your address book.  That takes care of the objection you list above,
most of the time.  If GIF's being part of a rendered message become
a significant spam indicator, it is likely that legitimate corporate
senders will avoid it.  That will leave a few exuberant parents, and
a whole lot of spammers, creating that style of message.  Given that
scenario, whitelisting is a reasonable solution.  MUA's could
eventually help the situation if they discouraged putting inline
images in a message.

I guess that bottom line is that I don't see any problem for the
average business email user.  Business newsletter senders will adapt
as necessary - they always do.  Individual home users will do
whatever their ISP's force on them.  I doubt that ISP's will care to
engage in an arms race that forces them to do OCR on every piece of
spam they receive.  SpamAssassin during SMTP is one thing, OCR raises
the ante and threatens the affordability of ordinary email.  Spam
becomes a credible DDoS.

-- 
Seth Goodman


More information about the SpamBayes mailing list