Summary
- ETH Zurich researchers modified an AI model to solve Google’s reCAPTCHAv2.
- The You Only Look Once (YOLO) picture processing model was adapted to pass the CAPTCHA test every time.
- The team trained the model on images of objects typically used in reCAPTCHAv2, like cars and traffic lights.
- Testing showed the modified AI model could even bypass more sophisticated CAPTCHAs with features like mouse tracking or browser history.