Scholarlink

Using CAPTCHA helps shield computers from malevolent bots. The vulnerability of modern text-based CAPTCHAs has been exposed by the emergence of deep learning techniques. Since image-based CAPTCHAs need a lot of labour to create, image-based visual reasoning is starting to emerge as a new direction in this progression. The Visual Turing Test (VTT) CAPTCHA was just released by Tencent. This appears to be the first time the visual reasoning method has been used. Other CAPTCHA service providers (such as Dingxiang, NetEase, and Geetest) also began to create visual reasoning frameworks of their own to combat robots. So, it makes sense to pose this fundamental query: Are CAPTCHAs that rely on visual reasoning as safe as their designers believe? In this paper, we introduce the first method for solving visual reasoning-based CAPTCHAs. On the VTT CAPTCHA, our modular strategy and holistic approach yielded overall success rates of 88.0% and 67.3%, respectively. According to the research, visual reasoning CAPTCHAs are not as secure as previously believed. Attempts to apply difficult, original AI tasks to CAPTCHAs have not been successful thus far. We also provide some recommendations for better secure, visually engaging CAPTCHAs based on our attacks.

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

Revolutionizing Web Security: The Efficiency of a Reusable CAPTCHA Security Engine

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Yamsani Vaishnavi,
  • Touria Tanazzum,
  • Hari Bharadwaj,
  • S. Ranjithreddy,
  • Kottu Santosh Kumar,
  • Saroja Kumar Rout

摘要

Using CAPTCHA helps shield computers from malevolent bots. The vulnerability of modern text-based CAPTCHAs has been exposed by the emergence of deep learning techniques. Since image-based CAPTCHAs need a lot of labour to create, image-based visual reasoning is starting to emerge as a new direction in this progression. The Visual Turing Test (VTT) CAPTCHA was just released by Tencent. This appears to be the first time the visual reasoning method has been used. Other CAPTCHA service providers (such as Dingxiang, NetEase, and Geetest) also began to create visual reasoning frameworks of their own to combat robots. So, it makes sense to pose this fundamental query: Are CAPTCHAs that rely on visual reasoning as safe as their designers believe? In this paper, we introduce the first method for solving visual reasoning-based CAPTCHAs. On the VTT CAPTCHA, our modular strategy and holistic approach yielded overall success rates of 88.0% and 67.3%, respectively. According to the research, visual reasoning CAPTCHAs are not as secure as previously believed. Attempts to apply difficult, original AI tasks to CAPTCHAs have not been successful thus far. We also provide some recommendations for better secure, visually engaging CAPTCHAs based on our attacks.