A Proposal of New Reading Text CAPTCHA using Random Dot Patterns

Abstract

According to the growth of troubles caused by malicious programs called bots, CAPTCHA (Completely Automated Public Turing test to tell Computers and Humans Apart) comes to be an important role in the current information society. CAPTCHA identifies bots from legitimate human users by requiring some questions that are easy for humans to solve but difficult for bots. However, the progress of computing technology such as Optical Character Recognition (OCR) function. bots come to be able to solve present CAPTCHAs. In order to outcome such troublesome bots, more sophisticated CAPTCHA that requires the high cognitive ability of human beings is desired. In this paper, we propose a new CAPTCHA scheme that uses random dot patterns. Human beings can recognize a moving figure filled by a random dot pattern from a background that is filled by another random dot pattern; however, they lose the figure in the background when the figure pauses. Since image recognition by computer programs is usually carried out frame by frame, it is hard for bots to recognize such a moving figure filled by a random dot pattern from another random dot pattern background. The proposed CAPTCHA scheme exploits this characteristic of random dot patterns. This CAPTCHA requires users to answer a text filled by a random dot pattern that is moving on a background also filled by another random dot pattern. Several experiments were carried out to confirm that the proposed CAPTCHA scheme has enough resistance against bot attacks using representative image recognition methods. Results of the experiments showed that the image recognition methods cannot find out the answer text. Other experiments were also carried out to evaluate the usability of the proposed CAPTCHA scheme. The system usability scheme was adopted in the experiments. The obtained score was 89 that means this CAPTCHA scheme has high usability. These results showed that the CAPTCHA scheme is usable enough and has enough resistance against bot attacks.