Home New Technology New Technology – Algorithm can fool CAPTCHA checks

New Technology – Algorithm can fool CAPTCHA checks


Researchers have developed an algorithm that can now outsmart websites that use CAPTCHA challenge-response tests for their security. It does this by imitating how the human brain responds to these tests’ visual clues.

You may have seen these CAPTCHA tests, which challenge you to prove you’re human by inputting combinations of letters and numbers – especially when you try to log in a website.

The algorithm is capable of identifying letters and numbers from their shapes, like people can.

Vicarious – an AI company funded by Amazon founder Jeff Bezos and Facebook’s Mark Zuckerberg –¬†created the algorithm.

How CAPTCHA works

The CAPTCHA test (which stands for “Completely Automated Public Turing test to tell Computers and Humans Apart”) keeps people from using automated bots to set up fake accounts on websites.

When logging into a website, users prove they’re human by solving visual puzzles, which usually involves identifying distorted letters, numbers, symbols, or objects. Automated bots typically struggle to pass these tests.

In fact, Google says that its reCaptcha test is so complex that even humans can only solve it 87% of the time.
Vicarious, however, says their algorithm can recognize these distorted letters and digits from images.

Neural networks

To get machines to identify images, computer scientists use neural networks, which are large networks of computers trained to solve complex problems.

A neural network contains hundreds of layers, whose design is inspired by the human brain. Each of these layers examines a different part of the problem. Eventually, the network combines answers from all the layers to produce one final result.

The problem is that humans need to “teach” neural networks by inputting thousands of pre-labeled images. It’s clearly a monumental project.

To solve this problem, Vicarious developed Recursive Cortical Network (RCN), a software that mimics human brain processes. It does this while requiring less computing power than a neural network.

Vicarious has been developing algorithms which examine pixels in an image to see if they match an object’s outline.

Success rates

In 2013, Vicarious said they beat CAPTCHA tests from Google, Yahoo, PayPal and Captcha.com with a 90% accuracy.

Since then, CAPTCHA designers have made their tests more difficult to complete. And yet, Vicarious said in their research that the algorithm can now pass Google’s reCaptcha test 66.6% of the time.

The RCN software was also able to solve reCaptacha tests from BotDetect at a 64.4% success rate. For Yahoo and PayPal, it succeeded at 57.4% and 57.1% of the time, respectively.