人工免疫算法在反垃圾邮件技术中的应用

【作者】 管琴琴；

【导师】 郝玉洁；

【作者基本信息】 电子科技大学 ， 计算机应用， 2008， 硕士

【摘要】 人工免疫系统是当前计算智能领域的新兴研究热点,其应用领域包括工业控制、模式识别、经济管理、计算机网络安全、反垃圾邮件等。近年来,由于生物免疫系统集成了邮件过滤器必需的优良特性,而人工免疫系统能够较好地从生物免疫中继承、发展这些特性,优于其他智能系统,体现出人工免疫原理与反垃圾邮件技术的结合的天然优势。将人工免疫原理及其算法应用到邮件过滤技术中能够提供比传统方法更为广泛的多样性、模式识别能力、自适应性和容噪性。人工免疫算法逐渐成为反垃圾邮件技术中的应用热点。本文以人工免疫系统原理和算法为研究对象,探索人工免疫算法在反垃圾邮件技术中的应用。首先对反垃圾邮件技术以及人工免疫系统原理进行总结归纳,并对电子邮件的基本理论、人工免疫的生物学背景和传统的反垃圾邮件技术进行了简要的介绍。以一个客户端的垃圾邮件过滤系统为平台,设计人工免疫原理的垃圾邮件过滤器,并详细研究与分析了人工免疫算法和IP-ISF算法在其中的应用。并进一步深入研究IP-ISF算法在检测效率稳定性方面地问题,提出了改进方案,最后通过改进算法的实现与测试结果分析验证了算法改进的有效性。最后提出了基于人工免疫算法的Outlook客户端垃圾邮件过滤插件的设计方案。本文的应用研究成果对人工免疫算法在其他应用领域,如垃圾短信的过滤等提供了借鉴和参考。本文的研究重点是人工免疫算法及在垃圾邮件过滤中的应用,给出了基于人工免疫原理的邮件过滤器的设计与实现机制,对IP—ISF免疫算法的应用进行了研究,详细分析了邮件社区基因库、特征检测器等算法模块的应用与实现。对基于人工免疫原理的邮件过滤器的设计和算法流程进行了探索,对实验结果进行了对比分析。本文的研究难点是,由于动态数据环境下IP-ISF垃圾邮件分类算法的正确检测率的不稳定性,既需要针对新特征类型提高初次检测的效率,又应针对旧有特征保存已有的优势基因,所以提出免疫优势位点演化来提高新生成特征检测器的检测效率,增强演化的方向性,提高准确识别的稳定性。并利用公共语料库进行测试,通过测试结果证明改进算法有效地增加了动态数据环境下正确检测率的稳定性,验证了本文所作的研究工作的有效性和可行性。更多还原

【Abstract】 Artificial immune system is the emerging field of computational intelligence research, its applications include Industrial Control, Pattern Recognition, economic management, computer network security, anti-spam, etc.. In recent years, Artificial Immune System which is inherited from the biological immunity, has developed it’s characteristics, and superior to other smart systems. It has shown the natural advantage of the combination of the technologies of the embodied artificial immune and anti-spam. Artificial immune principle and it’s application to e-mail filtering algorithm technology can provide more extensive chance than the traditional method, for instance, diversity, pattern recognition capability, adaptability and the capacity of noise defence. Artificial immune algorithm gradually become hot spots in anti-spam technology.Based on artificial immune system theory and algorithms’s study, this paper will explore AIS and immune algorithm in the anti-spam technology applications. First, anti-spam technology and artificial immune system principle summarized, then the e-mail’s basic theory, artificial immune biology background and the traditional anti-spam technology for a brief introduction. To introduce a client of spam filtering system as a platform, deeply research the design principle of artificial immune spam filters, and detailed study and analysis the artificial immune and IP-ISF algorithm in which the application uesd into. And further study of artificial immune algorithm improvements, through improved with the realization of the test results to improve the effectiveness of verification algorithm. Finally, present the proposed algorithm based on artificial immune Outlook client spam filtering plug-in’s design. In this paper, through the application of research results, artificial immune algorithm in other application areas, such as SMS spam filters provide reference.This study focused on artificial immune algorithm in the spam filters and the application that is based on artificial immune filters principle. The most important part is the design and implementation of immune algorithm and IP-ISF, which is a detailed analysis e-mail community algorithm, in the application of the Anti-spam system. Based on artificial immune’s principle, the experimental results are compared and analyzed with many other methods. In this study provide a good way to, for dynamic data environment correct detection rate’s stability, enhance the evolution of the rules of the new generation Detection efficiently, thereby enhancing the stability of accurate identification. Corpus tested on the platform mentioned before. And the work in this paper is a good case for other compuer applications.更多还原