Automatic Construction of Keyword Dictionary to Enhance the Security of Image-based CAPTCHA
Yeonchan Ahn, Yoo-Sung Kim · 2011
Previous image-based CAPTCHA systems proposed to improve the security of text-based CAPTCHA have some drawbacks; they are vulnerable against automated program's dictionary attack using color information of images because their keyword dictionary is static and they require some manual tasks to construct keyword dictionary so that they might have obstacles for being used in real Internet applications. In order to solve these problems we suggest an automatic construction method of keyword dictionary for image-based CAPTCHA, which extracts keyword candidates from web documents in the Internet and apply filtering processes to select appropriate keywords among them. The proposed automatic construction method of keyword dictionary can have not only dynamic property because it uses the Internet web document updated frequently but also high accuracy because it selects only appropriate keywords for image-based CAPTCHA using the filtering processes. Furthermore, our study suggests a guideline for choosing proper topics of web documents for automatic construction of keyword dictionary for image-based CAPTCHA by comparing effectiveness of the proposed method against several documents of different topics. And we proved that security level is improved by comparing the success rates between ours and the previous one against random guessing attacks.