Evaluating robustness of Arabic CAPTCHAs

摘要

CAPTCHAs are applied on websites to differentiate between human users and automated programs, which indulge in spamming and other fraudulent activities. Since there are many websites that provide services in Arabic language, Arabic CAPTCHAs have been proposed by a number of studies. These CAPTCHAs rely on the distortion of text images rendering them unrecognizable to essentially pattern recognition techniques. However, the segmentation resistance evaluation against the proposed Arabic CAPTCHA schemes has not yet been done. Thus, this paper evaluates the robustness of the state of the art Arabic CAPTCHAs. Specifically, we showed that a number of proposed Arabic CAPTCHAs could be broken with an acceptable success rate. Moreover, a set of recommendation is derived in order to guide the design of robust Arabic CAPTCHAs.