基于免疫算法和神经网络的新型抗体网络研究

【作者】 陈科；

【导师】 许家珆；

【作者基本信息】 电子科技大学 ， 计算数学， 2006， 硕士

【摘要】 本文涉及的课题是“基于免疫算法和神经网络的新型抗体网络研究”,人工免疫是当前计算智能领域的新兴研究热点.本课题以人工免疫系统和神经网络为研究对象,并依托四川省科技厅基础项目“智能入侵检测系统的关键技术研究”(项目编号:04JY029-017-1) ,将人工免疫原理和神经网络技术相结合,提出了应用于入侵检测系统的新型抗体网络,为该项目在工程领域中的进一步研究、应用提供理论依据.近年来,受生物系统启发而设计出来的智能算法越来越受到人们重视.免疫算法、神经网络、遗传算法并称为当今三大仿生算法.在各学科相互交叉相互渗透的今天,将免疫原理与神经网络技术结合,既弥补了神经网络收敛速度慢,容易陷入局部最优解的弊端,又增强了免疫算法的分布性和自适应性.第一代安全保护技术采用认证、授权、访问控制、防火墙和加密等外围防卫机制来抵御入侵行为,但对于规模迅速增大,日渐复杂的网络系统已经力不从心.入侵检测作为第二代网络安全技术,受到了国内外的广泛重视. IDS(Intrusion Detection System)作为一种积极主动的安全防护技术,提供了对内部攻击、外部攻击和误操作的实时保护,在网络系统受到危害之前拦截和响应入侵.本文介绍了免疫算法和神经网络技术的概念、原理及其在入侵检测系统中的应用,然后对比了两者在应用中的优缺点,最后通过对人工免疫系统和人工神经网络的比较,提出了新型抗体网络模型.本文的创新点在于提出了一种基于免疫算法和神经网络的新型抗体网络;将免疫算法应用于入侵检测系统,充分利用与神经网络协同工作的互补性来完成入侵检测任务,解决了传统入侵检测系统的自学习功能差的缺陷.将新型抗体网络模型应用于大型网络的入侵检测任务,具有良好的可扩充性.论文还阐述了如何引进BP神经网络自学习能力,针对已有的抗体网络模型进行改进.通过对网络数据集的测试,对模型进行完整的分析和测试,来检验模型的可靠性和高性能.测试结果表明该算法相对于传统抗体网络,其检测效率得到了明显的改善.更多还原

【Abstract】 Artificial Immunology is a new research field in computing intelligence. This paper uses Artificial Immune System and Neural Network as the research object. We use artificial immune principles to design a novel antibody network which based on the Immune Algorithm and Neural Network, and use these as in-depth research of the application of our achievement to engineering.In this paper, we built the novel Antibody Network which based on the two intelligence algorithms and reach the application of its use in the intrusion detection system. Furthermore we create an ideal antitype system from it.We first introduce the current research status of Artificial Immune System and Neural Network technique. then we simply present the basic theory of the Antibody Network , the biologic background of Artificial Immune System and some traditional Antibody Network technique. A new intrusion detection algorithm combines with Immune Algorithm and Neural Network technique was provided.This paper designs an Antibody Network base on Immune Algorithm and Neural Network used for Intrusion Detection System. The Antibody Network system combines network-based intrusion detection functions. It can be used to protect large area network and has relatively good expansibility. This paper also discusses the theory of the Antibody network how to implement of the Antibody Network with BP neural network in intrusion detection. Because of the traditional Antibody network weakness in learning time and performance, this network is designed. The algorithm has been tested on a network data set. The result shows that it had much better performance than traditional Antibody Network更多还原