Researchers have developed a method to bypass any captcha
The familiar website security test has given way to artificial intelligence, jeopardizing its effectiveness. An AI-based algorithm developed by an international team of researchers traverses CAPTCHAs using a self-learning neural network. The method has not yet received its own name. Unlike the time it takes for standard captcha input, the new algorithm is faster: artificial intelligence completes the task in 0.05 seconds. The CAPTCHA test is an empirical Turing test to distinguish between robots and humans. Its purpose is to identify between a person and a machine performing actions on the site. For the first time, computer captcha appeared in 1997. CAPTCHA validation tasks are used to protect many resources, including such giants as Wikipedia, the eBay platform and the official websites of large companies. To solve the question of how to bypass the captcha, the developers used a self-learning system without a teacher in the algorithm (the so-called generative adversarial network). The technology provides for the presence of two oppositely tuned neural networks: one generates many training samples, and the other analyzes them for authenticity. Thus, the researchers used multiple versions of indistinguishable CAPTCHAs to train the algorithm, while tweaking the AI to make the method effective.  The developers themselves noted that the use of a generative adversarial network has reduced the number of training samples to about 500, while without using the self-learning method, their number could reach several million. They also called the captcha bypass algorithm a visual way to show the vulnerability of a popular validation test in the event of an attack using artificial intelligence technology. The developers of the method practically applied it on several dozen of the most common variants of a computer test, with which the algorithm successfully coped. Despite the development of a neural network capable of bypassing CAPTCHAs, the verification computer test has a chance of rehabilitation. A few months ago, Google introduced the reCAPTCHA 3 technology, which does not require mandatory text entry, image recognition or other actions. The system is active in the background without direct human intervention. The algorithm takes into account the user's online behavior over a certain period, usually several days. The collected information is used by the reCAPTCHA technology to distinguish the actions of robots from the behavior of users. By the time of the start of work, the accuracy of the algorithm, according to the developers, reaches 99%. The Topic of Article: Researchers have developed a method to bypass any captcha. |
