Artificial Intelligence outsmarts CAPTCHA security
New algorithm uses human visual intelligence to fool the security system
Vicarious, an AI firm based in California, has created an algorithm which can beat CAPTCHA by mimicking the way the human brain processes images and visual clues.
CAPTCHA, which stands for Completely Automated Public Turing test is used to tell Computers and Humans Apart, was first introduced in the 1990's. Its purpose is to prevent people from using automated bots to set up fake accounts on websites and works by the user solving visual puzzles like recognising distorted letters, digits, symbols or objects. This is because algorithms find it difficult to both match image recognition to semantic meaning and to recognize familiar images that have been distorted.
The researchers at Vicarious claim their software, Recursive Cortical Network (RCN), however, can outsmart CAPTCHAs with minimal training. In fact, previous AI programs required 50,000 times more training than the RCN.
Unlike other processes which use neural networks that require extensive training to solve problems, the RCN works by mimicking how people's visual cortex processes pictures. According to their research published in the journal Science, the algorithm recognises contours, edges, shapes, and textures of an image, and then analyses the pixels to see if they match the outline of an object.
The research revealed that the algorithm was able to accurately guess a CAPTCHA image 66% of the time and can correctly guess an individual character with 81% accuracy.
"We're not seeing attacks on Captcha at the moment, but within three or four months, whatever the researchers have developed will become mainstream, so Captcha's days are numbered," Simon Edwards, a cyber-security architect for data cyber-security firm Trend Micro Europe, told the BBC.