Scholarlink

CAPTCHA (Completely Automated Public Turing test to tell Computers and Humans Apart) is an automated verification mechanism designed to distinguish between human visitors and automated systems as a vital component for cyber security. However, current CAPTCHA recognition technologies still struggle to keep pace with the sophistication CAPTCHA generation algorithms. This paper introduces a novel approach to CAPTCHA recognition, utilizing the Vision Transformer (ViT) and enhancing it with two key optimizations: the Permutation Visual Model, which improves the model’s spatial understanding, and Transfer Learning for the Vision Transformer, which accelerates the model’s adaptation to new tasks. The method was rigorously tested against a diverse set of practical data, simulating real-world scenarios. The results are promising, with an accuracy rate of over 95% across multiple websites, indicating a significant advancement in CAPTCHA recognition technology.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

A High Accuracy Text CAPTCHA Recognition Approach Through Opertimized Vision Transformer

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Wei Hao,
  • Shoulai Shang,
  • Yepeng Zhang

摘要

CAPTCHA (Completely Automated Public Turing test to tell Computers and Humans Apart) is an automated verification mechanism designed to distinguish between human visitors and automated systems as a vital component for cyber security. However, current CAPTCHA recognition technologies still struggle to keep pace with the sophistication CAPTCHA generation algorithms. This paper introduces a novel approach to CAPTCHA recognition, utilizing the Vision Transformer (ViT) and enhancing it with two key optimizations: the Permutation Visual Model, which improves the model’s spatial understanding, and Transfer Learning for the Vision Transformer, which accelerates the model’s adaptation to new tasks. The method was rigorously tested against a diverse set of practical data, simulating real-world scenarios. The results are promising, with an accuracy rate of over 95% across multiple websites, indicating a significant advancement in CAPTCHA recognition technology.