摘要： 验证码图片是论坛类网站用以阻止自动化程序恶意行为的重要人机区分技术，其设计和实用安全性直接涉及到互联网的正常使用。为研究国内验证码实用安全性，设计实现一种验证码识别算法模型，对国内的论坛验证码类型进行了实验和分析。实验结果表明，实用中的验证码识别率通常在50%以上，某些甚至达到100%，难以起到对自动化程序的阻碍作用。从实用角度将验证码分为4类，发现最佳的一类基本是空白的，表明验证码实用技术还应做较大改进。

关键词: 验证码, 自动识别, Internet安全

Abstract: CAPTCHA image is the major technique used by forums to tell computers and humans apart, so that its security is of importance in internet usage. To learn about domestic CAPTCHA images, this paper proposes and implements a model of CAPTCHA image recognition algorithm, and tests almost all image types used in domestic forums. Experiment results indicate that a large part of CAPTCHA images have a recognition rate above 50%, and some with 100%, both of which can hardly block automatic machines. In terms of its practicality, this paper further classifies them into four categories, only to find the optimal category is blank, which means CAPTCHA image technology needs great enhancements.

Key words: CAPTCHA, automatic recognition, Internet security

中图分类号: 

• TP309