OT: captchas

[Frank Wiles]

On Fri, 15 Apr 2005 17:03:46 -0500
Tim reid <bewkard at gmail.com> wrote:

> I have a question for the CS majors ;)
> 
> Captchas
> http://www.captcha.net/
> http://en.wikipedia.org/wiki/Captcha
> 
> Has anyone ever seen any writing on the idea of caching the different
> captcha images, manually decoding them and storing the "meaning" of
> the captcha in a database.  Then cross referencing the "meaning" of
> the captcha with the image filename.  That way you would have the
> ability to have a bot load a page/service protected by a captcha, read
> the image filename, and decode the captcha.
> 
> Is this a valid idea?

  It would be valid, but any captcha implementation worth it's salt
  doesn't use flat file images.  It generates a random name for the
  image and serves it up to the client. 

  Here is how it works: 

  1) Choose random captcha that happens to say "FooBar" which is in
     foobar.jpg. 

  2) Tell browser to load /images/AlkjsdfH293sdfhjh2234kjh.jpg 

  3) Have a system in place that, in the background, serves up
     foobar.jpg when asked for /images/AlkjsdfH293sdfhjh2234kjh.jpg 

  This keeps bots like you were thinking from working. Because each
  time the filename is different.

  P.S. FYI I'm not a CS major, not that it would have helped anyway
       as this isn't something they are going to teach you in school. :)

 ---------------------------------
   Frank Wiles <frank at wiles.org>
   http://www.wiles.org
 ---------------------------------