AI was the first to trick Microsoft’s captcha. After he was trained to automatically remove excess noise, the tool’s efficiency increased to 90%.
Researchers at F-Secure learned how to use AI to bypass captcha on Microsoft Outlook. In their text, they explained that the biggest difficulty for them was the correct definition of the text and the imitation of “live” typing. However, the rest of the instrument may pretend to be a normal user.
At first, AI could only detect 22% of the characters in a standard captcha. In the process, the researchers found that artificial noise makes it difficult to identify letters. To counteract this effect, they tuned the algorithm to automatically remove noise. After that, AI was able to circumvent more than 90% of captchas.
F-Secure in 2013 developed software that circumvented Turing’s automated tests on Google, Yahoo, and PayPal. It has demonstrated efficacy greater than 90%.
“We talked about this last year, let’s say it now: text captcha has long been ineffective and easy to get around. New interesting patterns have appeared on the market, but when they begin to get around – this is just a matter of time. We do not say that captcha is useless, they simply should not be considered as a solution that will save you from everything, ”noted F-Secure.