Reported by sergioneves on 2013-05-15 10:09
There is a new way of providing CAPTCHAs, which is based on the
introduction of randomly distributed text nodes on the DOM of web pages, and employment of CSS to arrange these nodes (characters)
on the screen.
One known example of the implementation of this type of CAPTCHAs is on Packt Publishing book seller web site, where the following link leads to a particular example:
http://www.packtpub.com/support/10503
If you analyse the CAPTCHA, you'll see that the order of characters announced by any screen reader (in particular NVDA) using the virtual cursor is not equal to the order of characters that really appears on the screen. I didn'tt analyse it thoroughly, but
I think the order is not coincident due to the CSS applied to text nodes.
This is one case where it's important to access and explore the web page with another ways other than the virtual cursor and probably to use low-level graphics functionality to inspect web page content; otherwise, I don't know any other way to discover the order of characters on these CAPTCHAs.
I think it's a long-term problem that is worth researching.