非粘连字符型验证码的安全性研究

发布时间：2018-06-09 13:22

本文选题：网络安全 + 文本型验证码　；参考：《北京交通大学》2017年硕士论文

【摘要】：随着互联网技术的高速发展,网络的安全性问题日益突出。验证码(CAPTCHA)技术能够判断登录用户是否是计算机模拟的虚拟用户,有效提高了网络安全。目前使用最为广泛的是文本型验证码。通过对文本型验证码识别技术的研究,可以及时发现验证码设计方案中存在的漏洞,并对其设计方案提出改善建议,从而提高验证码的安全性,进而提高网络安全。通常情况下,我们认为人眼识别率大于90%,机器识别率小于10%的验证码被认为是安全的。本文采用数字图像处理与模式识别等相关理论和方法,研究文本型验证码的计算机识别方法。本文主要工作和成果如下:(1)针对六种规整型字符验证码图像,给出了一种基于边缘检测的连通域分割字符的算法,建立了包括数字、英文字母和部分汉字在内的1384幅图像的模板库,利用了模板匹配法进行识别,达到了 90%以上的正确率;(2)针对非规整型字符验证码图像,本文利用了 Chars74K数据集和从图像中分割的字符制作训练样本,采用了 BP神经网络进行训练识别,达到了 90%以上的识别正确率;(3)分析了验证码识别中的关键技术,给出了一种验证码识别的设计方案,并设计开发了一款跨平台的验证码识别软件,主要功能包括图像的预处理、去噪、边缘检测、字符分割和字符识别等。本文利用了改善的模板匹配法和BP神经网络对七种验证码图像进行了识别,均达到了较高的识别率,发现了七种验证码图像中存在的安全漏洞,并针对性地给出了几点提高验证码安全性的建议。
[Abstract]:With the rapid development of Internet technology, network security issues become increasingly prominent. CAPTCHA (CAPTCHA) technology can determine whether the login user is a virtual user simulated by computer and improves the network security effectively. At present, the most widely used is the text-based verification code. Through the research of the text type verification code recognition technology, we can find the flaw in the design scheme of the verification code in time, and put forward the improvement suggestion to its design scheme, thus enhance the security of the verification code, and then improve the network security. In general, we think that the verification code with a human eye recognition rate greater than 90 and a machine recognition rate less than 10 percent is considered safe. In this paper, the computer recognition method of text verification code is studied by using the theories and methods of digital image processing and pattern recognition. The main work and results of this paper are as follows: (1) aiming at six kinds of regular character verification code images, this paper presents an algorithm of character segmentation based on edge detection in connected domain. The template library of 1384 images, including English letters and some Chinese characters, is recognized by template matching method, and the accuracy is more than 90%. In this paper, we use Chars74K data set and character segmentation from image to make training sample, and use BP neural network to train and recognize. The recognition accuracy is more than 90%.) the key technology of verification code recognition is analyzed. In this paper, a design scheme of verification code recognition is presented, and a cross-platform verification code recognition software is designed and developed. The main functions include image preprocessing, denoising, edge detection, character segmentation and character recognition. In this paper, the improved template matching method and BP neural network are used to identify the seven kinds of CAPTC-code images, and the recognition rate is higher, and the security holes in the seven kinds of CAPTC-code images are found. Some suggestions to improve the security of CAPTCA are given.
【学位授予单位】：北京交通大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP393.08

【相似文献】