【作者】 吕英杰；

【导师】 李一军；

【作者基本信息】 哈尔滨工业大学 ， 管理科学与工程， 2007， 硕士

【摘要】 随着互联网技术的迅速发展,电子邮件作为互联网上使用最广泛、最受欢迎的服务之一,成为人们之间日常交流和沟通的重要手段。但是随着垃圾邮件的出现,这一快捷便利的通讯方式受到了前所未有的挑战。大量的垃圾邮件不但严重浪费了网络资源和人们的宝贵时间,而且还传播各种非法信息和病毒,给社会和人们生活造成危害,反垃圾邮件的相关研究也成为人们日益关注的热点问题。本文也正是在这种背景下对反垃圾邮件的相关方法进行研究。首先本文介绍了垃圾邮件的产生背景及其对社会的严重危害,表明反垃圾邮件的相关研究有着深远的社会意义和巨大的经济价值。随后介绍了反垃圾邮件的研究现状,主要包括政策、经济和技术等三方面的研究成果以及存在的缺陷。由于本文的研究是建立在垃圾邮件过滤技术的基础上,因此对相关的技术手段作了比较细致的介绍,为本文接下来的研究工作做好铺垫。在已有研究成果的基础上,本文提出了一种新的解决方案---基于全球IP信誉系统的垃圾邮件过滤方案。主要原理是通过对全球的IP地址进行信誉评估,并以电子邮件的发送源IP地址的信誉度为主要参考依据来辨识某邮件是否为垃圾邮件。这种方法与传统过滤技术相比,具有方便快捷、灵活性好和可靠性高等许多优点。在全球IP信誉系统的理论框架的基础上,本文采用BP神经网络技术来构建数据模型。通过分析系统的需求本文提取出几个对判别垃圾邮件具有重要影响的IP地址属性作为神经网络模型的输入,然后根据一些已知试验样本进行系统的训练及测试,最后得到高效可靠的网络模型用于评判IP地址的信誉等级。由于本文建立的全球IP信誉系统并不单纯是为了进行IP地址的信誉评价,而且还应该能从该框架体系中提取出垃圾邮件源IP的典型特征,这样才能更方便的结合其他的反垃圾邮件技术来进行综合评价。为此本文在构建神经网络模型的基础上作了进一步研究,通过样本试验从该模型中提取出了垃圾邮件源IP和非源IP的一些典型属性特征,作为辨别垃圾邮件的规则模式,为进一步研究反垃圾邮件相关技术提供了非常有价值的信息。更多还原

【Abstract】 With the rapid development of Internet, Email, as one of the most widely used and popular service, has become a ubiquitous and arguably essential means of communication. But now this means of communication suffers challenge from Email spam. Larger number of spam not only waste network resource and people’s time but spread various harmful information and network virus. So anti-spam research has become a hotspot issue.In this paper, we first introduce the background of the Email spam and the severe harm for the society to indicate that anti-spam research has far-reaching meaning and value. Then we introduce the actuality of anti-spam research, including policy, economy and technology. Since this research is built on email spam filtering technologies, we introduce interrelated technology in detail.Base on existed research fruit, we develop a new solution—global IP Reputation System. The main principle is that evaluating the reputation of global IP addresses, we can regard the reputation value of E-mail’s IP address to distinguish whether a Email is Email spam. Compared with traditional filtering technology, this approach has the strongpoint of convenience, agility and reliability.This paper adopts BP neural network technology to establish data model. By analyzing system demand we choose several IP address attributes that have the great value in distinguishing spam as the input of neural network model. Then using some existed experiment data to train the model, we finally get reliable network model to judge IP address reputation degree.Global IP Reputation System is not only for IP address reputation evaluation, but can extract the IP characteristics involved in spam in order to integrate other spam filtering technologies easily. So based on the neural network above model, we do some further research to extract some rules to distinguish whether one email is spam or not, offering some valuable information for further research.更多还原