Scholarlink

CAPTCHAs are widely deployed as a Turing test to distinguish computers from humans. Slide puzzle CAPTCHAs are becoming more and more popular due to user-friendliness and diverse defense mechanisms. At present, there are few studies on the security of slide puzzle CAPTCHAs, and the defense mechanisms of slide puzzle CAPTCHAs studied in the past were relatively simple. The existing slide puzzle CAPTCHAs use rich notch shapes, add complex interferences to the background images, detect the characteristics of the sliding trajectories in the two-dimensional space. These advancements present challenges to existing attack methods. In order to test the security of the existing slide puzzle CAPTCHAs, six kinds of very popular slide puzzle CAPTCHAs including Tencent CAPTCHA, GeeTest CAPTCHA v3, Shumei CAPTCHA, NetEase CAPTCHA, GeeTest CAPTCHA v4 and Dingxiang CAPTCHA are selected. Firstly, the Notch Detection Dataset and the Two-Dimensional Sliding Trajectory Dataset of each type of CAPTCHA are constructed for attacking slide puzzle CAPTCHAs. Secondly, this paper is the first to apply generative adversarial network (GAN) to attack slide puzzle CAPTCHAs. Finally, six kinds of CAPTCHAs are broken with the success rates close to 100%, which proves the effectiveness of this attack method. Experimental results show that the security of slide puzzle CAPTCHAs faces great challenges. The research work of this paper provides a novel idea for the security analysis of slide puzzle CAPTCHAs in the future.

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

SPC-GAN-Attack: Attacking Slide Puzzle CAPTCHAs by Human-Like Sliding Trajectories Based on Generative Adversarial Network

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Rui Guo,
  • Ke Yu,
  • Enbo Yu,
  • Qianqian Qiao,
  • Zhenyu Mao,
  • Haizhou Wang

摘要

CAPTCHAs are widely deployed as a Turing test to distinguish computers from humans. Slide puzzle CAPTCHAs are becoming more and more popular due to user-friendliness and diverse defense mechanisms. At present, there are few studies on the security of slide puzzle CAPTCHAs, and the defense mechanisms of slide puzzle CAPTCHAs studied in the past were relatively simple. The existing slide puzzle CAPTCHAs use rich notch shapes, add complex interferences to the background images, detect the characteristics of the sliding trajectories in the two-dimensional space. These advancements present challenges to existing attack methods. In order to test the security of the existing slide puzzle CAPTCHAs, six kinds of very popular slide puzzle CAPTCHAs including Tencent CAPTCHA, GeeTest CAPTCHA v3, Shumei CAPTCHA, NetEase CAPTCHA, GeeTest CAPTCHA v4 and Dingxiang CAPTCHA are selected. Firstly, the Notch Detection Dataset and the Two-Dimensional Sliding Trajectory Dataset of each type of CAPTCHA are constructed for attacking slide puzzle CAPTCHAs. Secondly, this paper is the first to apply generative adversarial network (GAN) to attack slide puzzle CAPTCHAs. Finally, six kinds of CAPTCHAs are broken with the success rates close to 100%, which proves the effectiveness of this attack method. Experimental results show that the security of slide puzzle CAPTCHAs faces great challenges. The research work of this paper provides a novel idea for the security analysis of slide puzzle CAPTCHAs in the future.