MapTCHA: A Novel CAPTCHA Leveraging AI Uncertainty to Combat Bots

Traditional CAPTCHAs leak user data and are costly to maintain. This paper introduces MapTCHA, a novel CAPTCHA that leverages the uncertainty of AI-powered computer vision in image interpretation to combat bots and spam. MapTCHA presents users with a mix of images containing AI-predicted objects (known positives, known negatives, and unknowns), asking them to identify correctly interpreted building outlines and other objects. User votes determine the truth of unknown images, providing new data sources for OpenStreetMap. The system uses the open-source AI-assisted mapping system fAIr for image recognition. Future plans include expanding to more objects and image types and integrating MapTCHA into various login systems.