网页信息监听分析系统的设计及实现.doc

Windows下网页信息监听分析 系统的设计与实现 摘要 随着我国互联网产业的迅速发展，网络信息安全和不良信息过滤已经成为了一个突出的技术和社会问题。设计实现了一个针对局域网和本机的网页信息监听分析系统，主要包含了网页监听、敏感文字探测、敏感图像探测、数据库存储四大模块。通过WinPcap和Libnids实现了网页信息的监控和监听。为了进行文字有害信息的准备过滤和鉴别，使用了基于正则表达式的关键词过滤技术。对于含有敏感关键词的一般文档，为了不将其错误过滤，提出了基于双词文法的文本分类模型。通过对正反两方面的文本的分词及词性标注，得到较为准确的双词文法的特征词列表，通过支持向量机，应用径向基函数核，对文本是否含有不良信息进行分类。在敏感图像探测上，综合了前人的工作，总结了一套完整有效且易于实现的色情图片检测算法，取得了85%以上的准确率。 关键词：网络监听,双词文法，正则表达式，支持向量机，皮肤检测 ABSTRACT With the rapid development of Chinese Internet industry, network information security and harmful information filtering has become an outstanding technical and social problem.?This design implements a web information monitoring analysis system consists 4 modules: a web page sniffer, sensitive word detection, sensitive image detection, database storage.?We use WinPcap and Libnids library to control and monitoring information on the website.?For the preparation of harmful information text filtering and identification, we use regular expressions based on keyword filtering technology. For general document which contains sensitive keywords, in order not to be an error filtering, we have proposed two-word grammar of text-based classification model.?Through both positive and negative annotation of the text word and part of speech tagging, we get more accurate two-word list of grammatical features of words. Through using the radial basis function kernel in support vector machine, we train a classifier to judge whether a document contains objectionable information or not.?In the sensitive image detection, we have consolidated the work of predecessors, summed up a complete set of effective algorithm to detect pornographic?pictures with over 85% precious which is easy to be implemented. 目录 1 背景及意义 4 2 系统简介 7 2.1整体架构 7 2.2 数据包捕获模块相关技术 9 2.2.1网络监听基本原理 9 2.2.2网络监听程序原理 12 3 算法设计 13 3.1敏感文本检测算法设计 13 3.1.1基于正则表达式的关键词匹配 14 3.1.2基于双词文法的文本特征提取 16 3.1.3双词文法的特征样本词对提取 18 3.1.4支持向量机作为分类器 20 3.2敏感图片检测算法设计 23 3.2.1原始图片的颜色空间转换(RGB-YCbCr) 24 3.2.2利用线性分类器找出所有皮肤像素 25 3.3.3计算图片的纹理特征(La