Captcha test 'cracked' by US

A US-based start-up claims to have broken security tests used to tell humans and computers apart online.

Vicarious said it had developed technology, based on the human brain, which could solve text-based Captcha tests 90% of the time.

A Captcha is a graphic or sound users must type on to a web page to prove they are human.

The company said its artificial intelligence software can also perceive images.

Artificial brains

The company said it had used its Recursive Cortical Network software to solve Captcha tests as a step towards thinking machines, not for nefarious purposes.

Vicarious hopes eventually to use the technology for robotics, medical image analysis, and online searching.

"The Vicarious algorithms achieve a level of effectiveness and efficiency much closer to actual human brains," Vicarious co-founder D Scott Phoenix said in a statement.

The artificial intelligence software can "think and learn like a human" by mimicking processes in the brain.

Facebook co-founder and Vicarious investor Dustin Moskovitz said that Vicarious was "at the forefront of building the first truly intelligent machines".

Security implications

In tests, the Recursive Cortical Network software solved 90% of Captcha tests used by Google, Yahoo, PayPal, and Captcha.com, Vicarious said.

Nevertheless, Captcha as a security mechanism was still valid, according to security experts.

The technology was originally developed as an anti-spam measure by a team at Carnegie Mellon University.

Computer scientist Luis von Ahn, who was part of the team that developed Captcha tests, said that it was difficult to verify Vicarious' results as the algorithm had not been made public.

Mr von Ahn said if necessary the Captcha test could be made stronger by increasing the distortion of Captcha text and images.

Paul Wood, a cyber-intelligence manager at computer security company Symantec, said the Vicarious announcement should be "no cause for concern for organisations developing Captcha technology".

However, Mr Wood added that technical and other weaknesses exist in some implementations of Captcha tests.

Audio and visual Captcha images may be difficult to distort further, as a number of users already find the tests difficult to transcribe, said Mr Wood.