A new algorithm can defeat CAPCHA test, making it more difficult for websites to tell if a human is using them or a bot
Until now, the text version of CAPCHA has been completely effective in controlling spam and malware bots. While CAPCHA can take any form, the text form was considered the most effective because the software has difficulty understanding warped, overlapping or camouflaged letters. Humans find it no problem at all to read any variation of letters, regardless how obfuscated they are.
According to Dileep George, Vicarious’ co-founder, the new algorithm can beat even Google’s CAPCHA, considered 90 percents safe.
According to George, he and his CEO Scott Phoenix believe that the new algorithm is leading the way to human-like artificial intelligence. Their program utilizes virtual neurons linked in a network designed to model the human brain. It enables their system to solve the problem in a similar way the human brain does.
Vicarious did not reveal details of their system, so the scientists who are experts in AI are not sure whether they created a leap in technology, or it is just one more system similar to others being currently deployed.
According to Selmer Bringsjord from the Rensselaer Polytechnic Institute, breaking text-based CAPTCHAs shows a high-level computer understanding of the letters. If
Vicarious can proof their claim, their achievement could be very significant. Vicarious wants to test its invention against some more Turing tests before going to the market with it. Phoenix says that they are focusing on solving the fundamental problems of artificial intelligence. Solving CAPCHA was a by-product.
Source: New Scientist