Video system designed to capture auto-challenge market

The people behind a new challenge/response test say their video-based system is both more legible to humans and trickier for computers to defeat.

NuCaptcha is the latest incarnation of the Completely Automated Public Turing test to tell Computers and Humans Apart, a particularly convoluted way of producing an acronym that sounds like “capture”. The idea is to verify a form has been completed by a human rather than a computer that is likely trying to disseminate spam.

Traditionally users had to type in a word that is originally displayed as an image. The problem with that method is that the better computers get at scanning the image, the more unconventional the typography has to be, to the point that some CAPTCHAs are difficult if not impossible for people to read.

NuCaptcha uses a video with scrolling text, the idea being that this increases the complexity of the data (making it harder to automatically read) while boosting the contrast between text and background. According to Leap Marketing Technologies, the company behind NuCaptcha, 99% of users get the text correct first time, compared with just 80% on a “competitor” system.

Speaking of the CAPTCHA, if you’ve ever been confronted with one featuring two words, the chances are you’ve unwittingly aided Google with its Book Search process. Since last year, when it bought out a company named reCAPTCHA, its been using the challenges to get human input into the book scanning process.

Whenever Google comes across words in books which its optical character recognition can’t recognize, it transfers the images into its CAPTCHA database. In some cases when users are presented with two words, only one is a genuine test that decided if they are verified as a human. If they get this right, their answer for the other word goes onto Google’s database. If enough people give the same solution, that’s taken as the correct reading of the word in the original printed document.