tag:blogger.com,1999:blog-1426281128647295476.post1321987429277369584..comments2010-03-03T02:21:53.471-05:00radihttp://www.blogger.com/profile/11459795773754260693radi@r-n-d.orgBlogger1125tag:blogger.com,1999:blog-1426281128647295476.post-8230417758170508682010-03-03T02:21:53.462-05:002010-03-03T02:21:53.462-05:00RE the Captcha issue: as far as I know, reCAPTCHA does not allow devs to specify their text - it&#39;s quite the opposite: reCAPTCHA is free because it shows users snippets of problematic digitizations of books, and uses the users&#39; answers to improve the OCR capabilities of the digitization. <br /><br />The issue described here is that there is a <i>file ID</i> (i.e. a different filename) for each image; however, even if reCAPTCHA used random GUIDs for the file names, you could still hash the image content and build the DB associating the hash with the answer...<br /><br />The real problem is the (small) number of different captcha challenges. A possible solution would be for reCAPTCHA to randomly modify the images (e.g. adding noise, warping the images, etc.) so to have a potentially infinite number of different captcha images.Gabriele Giuseppinihttp://www.blogger.com/profile/13958752731645145468noreply@blogger.com