Enhancing CAPTCHA security with YOLOv8: Deep learning for robust recognition and protection
Upendra Mishra, Pushpendra Kumar, Gagan Thakral, Bharti Chugh · 2025
Amid increasing cyber risks, CAPTCHA verification is a crucial protection tool that helps differentiate between genuine users and automated bots, strengthening online platforms against harmful attacks. Textbased CAPTCHAs are widely used and their effectiveness depends on posing a difficult task for bots while yet being understandable for human users. Utilizing advancements in deep learning and Computer Vision, it has become more feasible to develop models that are skilled at interpreting text-based CAPTCHAs. This study utilizes the YOLOv8 method, a cutting-edge object identification model, for recognizing CAPTCHAs using binary images as input. We adapt YOLOv8 to accurately recognize CAPTCHAs by repeatedly replicating the original input image and adding distinct binary images representing the characters in the CAPTCHA. Our method eliminates the requirement to separate CAPTCHAs into distinct character sets by utilizing the built-in features of YOLOv8 and its efficient design. The experimental results highlight the efficacy of our method in precisely identifying CAPTCHAs after thorough training and testing, showcasing its ability to improve online security measures against automated assaults.