Majority of the world’s most popular websites uses text captcha security system to protect their website. The new algorithm which is based on deep learning methods developed by computer scientist from three counties are now able to crack captcha that no other attack system is capable of. Moreover, this new artificial intelligence is efficient in recognizing captcha that it can solve it in 0.05 second by using desktop PC.

The key technique using is called “Generative Adversarial Network” or GAN. Under this technique, the researchers only need 500 genuine captchas to train a solver instead of manually tag captchas at previous stage. Also, since the new solver only require little human involvement, it can be easily adjust to crack new or modified captcha schemes. In the research, the program was tested on 33 captcha schemes and many of them are popular websites like eBay and Microsoft. The success and efficient in cracking captcha security system gives warning for both websites users and creators. Researches hope websites should consider multiple layers of security instead.

Source: https://www.sciencedaily.com/releases/2018/12/181205093701.htm