CAPTCHA Design Based on Moving Object Recognition Problem JingSong Cui, LiJing Wang, JingTing Mei, Da Zhang, Xia Wang, Yang Peng, WuZhou Zhang School of Computer Wuhan University Wuhan, Hubei, China E-mail: cuijs@qq.com Abstract—CAPTCHA is a test that can tell humans and computer programs apart automatically. The aim is to allow the server to identify the visitor is a human or a computer, and only provide services to human. It can improve the current server system and user information security. The static plane visual CAPTCHA based on OCR problems with the advantages of implementation and operation [1] become the mainstream of the current CAPTCHA technology application form. However, with a variety of targeted text segmentation technologies merging, such CAPTCHA based on OCR problems is faced with increasing security threats. In this paper, a new CAPTCHA based on the moving object identification and tracking problems is proposed, which is referred to biological motion vision model. An Innovative Single-frame Zero-knowledge rule is also put forward to make the CAPTCHA generation algorithm based on Edge Mutation. An attacker can log on the test service system, only after he solves the moving object recognition problem . Such animation CAPTCHA will be able to resist the attacks of all the static OCR technology, and resist the mainstream of attacks against the moving object detection. Keywords- Edge Mutation; Moving Object Recognition; Single-frame Zero-knowledge; CAPTCHA; Network Security I. INTRODUCTION In order to improve the security of the server and verify that the client request is submitted by individual users from online operations rather than malicious software, the academia proposed Completely Automated Public Turing Test to Tell Computers and Humans Apart (CAPTCHA) technology [2,3,4] , so that the server can automatically identify whether the visitor is a human or a computer. The static plane visual CAPTCHA based on OCR problems with the advantages of implementation and operation has become the mainstream of the current CAPTCHA technology application form. However, with a variety of targeted text segmentation technologies merging, such CAPTCHA based on OCR problems are faced with increasing security threats. CAPTCHA used by a number of websites at home and abroad including Microsoft MSN are attacked by programs. The network service systems protected in this way are increasingly being attacked by bots program. There is an urgent need for a new CAPTCHA which can not only improve the safety but also maintain the practicality . Therefore, in this paper a new animation CAPTCHA based on the moving object identification and tracking problems is proposed. An attacker can log on the test service system, only after he solves the moving object recognition problem. If an attacker can’t solve the moving object recognition problem effectively, he can’t continue attacking on the service system. II. RELATED WORKS CAPTCHA implementations are divided into three categories by the academia currently [5] : the visual programs based on OCR problems, visual programs based on non- OCR problems, non-visual programs. The visual programs based on OCR problems with the advantages of implementation and operation [1] are most thoroughly studied and widely applied. Means to achieve it as follows: When a visitor attempts to access Server, a picture containing a string of random numbers or characters is generated by server, and is added some interference. Server sends the picture to the visitor, and asks visitors to identify the words. The following is part of the OCR graphics CAPTCHA example: However, with the widespread use of OCR CAPTCHA, attack algorithms for it are continuously popping up. OCR recognition rate is getting higher and higher [6,7,8] . There are four typical attack methods. Statistical-based method has good robustness, better anti-noise and anti-jamming capability, while methods based on structural analysis are more sensitive to structural features and have better ability to distinguish similar words. Therefore people combine the two methods to form a third one. The third one is an important research direction [9,10] in CAPTCHA identification areas during recent years. The forth one is based on neural network. There are many neural network models, such as the BP network, Hopfield networks, self-organizing neural network model, etc. A lot of research results are achieved by using BP neural network method to validate CAPTCHA recognition in recent years [11,12] . Single character can be identified by the standard methods such as BP neural network with a high success rate [13,14] . For the above four kinds of attack methods, each has a number of concrete and feasible technical routes. Don’t repeat them here. Figure1. Easy OCR graphics CAPTCHA 158