Comments on: DR Raman explains CAPTCHA’s at Google Blog http://www.accessibilityblog.com/2006/11/29/dr-raman-explains-captchas-at-google-blog/ Raising the awareness of Web Site Accessibility and Marketing. Sat, 11 Oct 2008 06:50:05 +0000 http://wordpress.org/?v=2.6.2 By: Jesse the K http://www.accessibilityblog.com/2006/11/29/dr-raman-explains-captchas-at-google-blog/#comment-145525 Jesse the K Thu, 25 Oct 2007 11:22:58 +0000 http://www.accessibilityblog.com/2006/11/29/dr-raman-explains-captchas-at-google-blog/#comment-145525 Dan, you've given an excellent example of the Turing Test approach to filtering who gets in to a net service. But there's a deeper problem I encountered when administrating a UBB-based forum last year. There are genuine humans who are (presumably paid) to join services for the express purpose of spamming. Dan, you’ve given an excellent example of the Turing Test approach to filtering who gets in to a net service.

But there’s a deeper problem I encountered when administrating a UBB-based forum last year. There are genuine humans who are (presumably paid) to join services for the express purpose of spamming.

]]> By: Dan Culley http://www.accessibilityblog.com/2006/11/29/dr-raman-explains-captchas-at-google-blog/#comment-18351 Dan Culley Thu, 07 Dec 2006 20:22:55 +0000 http://www.accessibilityblog.com/2006/11/29/dr-raman-explains-captchas-at-google-blog/#comment-18351 Not only are image and (even) audio CAPTCHAs inaccessible, they are unnecessary. There is an entire universe of questions that are trivial for human beings to answer but pose significant difficulties for machines. Consider: <blockquote>Q: "What color are the teeth in your mouth when you are sad?"A: White.</blockquote> or <blockquote>Q: "If you put two apples next to two oranges and two rocks, then how many pieces of food are there?"A: Four.</blockquote> Now, I give spammers a lot of (begrudged) credit for their ingenuity in coming up with ways to break CAPTCHAs, but breaking these types of questions would require some pretty serious AI. I'm willing to bet if a spammer could answer these questions, he'd be selling the AI, not spamming. The bonus from using this type of method is that you don't need expensive backend components to generate the images or audio. And, of course, since it's text only, it's accessible. Not only are image and (even) audio CAPTCHAs inaccessible, they are unnecessary. There is an entire universe of questions that are trivial for human beings to answer but pose significant difficulties for machines. Consider:

Q: “What color are the teeth in your mouth when you are sad?”A: White.

or

Q: “If you put two apples next to two oranges and two rocks, then how many pieces of food are there?”A: Four.

Now, I give spammers a lot of (begrudged) credit for their ingenuity in coming up with ways to break CAPTCHAs, but breaking these types of questions would require some pretty serious AI. I’m willing to bet if a spammer could answer these questions, he’d be selling the AI, not spamming.

The bonus from using this type of method is that you don’t need expensive backend components to generate the images or audio. And, of course, since it’s text only, it’s accessible.

]]>