Google To Drop Text-Based CAPTCHA

gascaptcha

Google is to start replacing its CAPTCHA tests with a simple request for users to confirm they are not a robot. The company believes the new system will be easier for humans and harder for computers.

Until now Google, and numerous sites that use its CAPTCHA service, have asked web users to type two words shown in images of scans from books. Only one of these words is actually the test; the other is a cheeky way for Google to get help with words its system wasn’t able to automatically recognize after scanning.

Humans struggling to make out the words and the test thus becoming an irritation is one problem with the system. The other is that it’s no longer as secure as originally intended. Ironically Google’s own attempts to develop automated recognition for figuring out building numbers captured in shots for its Street View maps were so successful it was able to achieve 99.8 percent accuracy when it used the tool on its own CAPTCHA system.

Google now says it will roll out a revised system that, on the face of it, simply asks the users to tick a box next to the words “I’m not a robot.” What’s actually being tested is the time it takes the user to figure out what to do and move the mouse, plus the movement of the mouse throughout the process. The theory is that humans will do this in an unpredictable, imprecise manner, whereas computers trying to replicate the task will be too neat and predictable.

Users on mobile devices will carry out a separate test in which they are shown an image, then asked to pick the matching image (or images) from a selection of nine shown below. That’s the type of task that humans can do much quicker than a computer, partly because we have the ability to determine and match a few key features.

turkey_captcha

It’s not the end of the traditional text-based CAPTCHA however. That will remain as a back-up test in case the new test doesn’t produce a conclusive result.




Comments are closed.