Abstract

We present a Reinforcement Learning (RL) methodology to bypass GooglereCAPTCHA v3. We formulate the problem as a grid world where the agent learnshow to move the mouse and click on the reCAPTCHA button to receive a highscore. We study the performance of the agent when we vary the cell size of thegrid world and show that the performance drops when the agent takes big stepstoward the goal. Finally, we used a divide and conquer strategy to defeat thereCAPTCHA system for any grid resolution. Our proposed method achieves asuccess rate of 97.4% on a 100x100 grid and 96.7% on a 1000x1000 screenresolution.

 

Quick Read (beta)