Sunday, April 19, 2009

Upright Rotation CAPTCHAs

A Google R&D post describes a new type of CAPTCHA based on asking users to rotate a set of images to a natural upright position.

See: Socially Adjusted CAPTCHAs and the complete R&D paper What’s Up CAPTCHA? A CAPTCHA Based On Image Orientation.

After filtering out images that contain features easily recognizable by computers - such as faces, cars, pedestrians, sky, grass - performing this task successfully is easily done by people but not by computers.

Displaying three images and requiring the upright rotation to be within a 16-degree window (8-degrees on each side) results in random computer-generated guesses being correct in fewer than 1 in 10,000 tries.

An issue I see in the researchers’ reasoning is that a computer does not have to rotate the images to a naturally upright position to break this What’s Up CAPTCHA.

If a computer can detect the overwhelming vertical or horizontal lines or edges in an image the computer can rotate the image on the X or Y axis to an upright, upside-down, left-facing, or right-facing orientation. This would mean three images could be guessed correctly 1 in 64 (not 10,000) tries.