Research On The CAPTCHA Technology Of Consensus Warfare In Network

With the rapid development of Internet, network has brought great convenience to people’slife, but at the same time the network security problems are increasing rapidly.CAPTCHA, served as atool to distinguish computer and human, has been widely used for network security. At present, theresearch of CAPTCHA recognition is more and more mature, while segmentation of CAPTCHA isstarted relatively late and the characters variability is relatively large, and the research of segmentationis much more difficult than recognition.Thus, the study of complex segmentation of CAPTCHA hasbeen the most difficult problem.But it has a very important significance to enhance the security ofnetwork and prevent malicious attack from web sides.This article mainly analyzes the CAPTCHAs which are formed with connected characters. Withthe comparison of several previous algorithms, the research has a certain practical value and function.The main work and results are as follows:1) Aimed at the characteristics of connected characters, this paper put forward a segmentalgorithm of community division which is based on complex network and compared the results of thesegmentation to different numbers of connected characters. This paper draw a conclusion that thealgorithm is effective to segment these characters and the successful segmentation rate of this algorithmto the CAPTCHAs for Authorize,360buy, Tianya, Window Live and Taobao can reach98%,95%,71%,55%,33%, respectively. At the same time, the more connected characters are, the lower the successfulrate is, and the longer the segmentation time is.2) Aimed at the characteristics of the CAPTCHAs from Tianya, this paper proposed analgorithm based on Water Reservoirs. After comparing the several different algorithms, this paper gotfinal result of the segmentation with a successful rate of92%and it is very effective to segment theoverlaped characters or the pixels are too large between two characters. It’s also effectively resolved thelimitation of the segmentation of community division algorithm and reduced the segmentation time.3) Aimed at the recognition of CAPTCHAs, this paper proposed the33features extraction methodand used the C-SVC to recognize it. The experiment result showed that the successful rate of thismethod can reach93%to Tianya. Finally, this paper introduced the implementation of CAPTCHAssystem, which including acquisition, pre-processing, segmentation and recognition.