云环境下的入侵检测关键技术研究

发布时间：2018-03-16 21:16

本文选题：云计算　切入点：Hadoop　出处：《华北电力大学》2017年硕士论文　论文类型：学位论文

【摘要】：随着计算机网络的普及和信息技术的发展,作为一项新兴技术,云计算越来越多地应用于社会生活的各个领域。云计算强大的计算能力和存储资源为众多企业和个人带来便利,而与此同时,它巨大的经济价值使其成为众多黑客的攻击目标。如今,云安全问题已成为云计算发展过程中面临的头号挑战。面对新型的安全威胁,如何保障云环境的信息安全,是当前态势下亟待解决的课题。入侵检测系统是动态网络安全防护的重要手段,基于云计算的入侵检测系统已成为云安全体系重要的组成部分,云环境下的入侵检测系统也是入侵检测和云计算领域新的发展方向。本文分析了入侵检测技术以及云计算的特点,结合云环境的安全需求,基于通用的入侵检测标准模型CIDF,设计了一个云环境下的入侵检测系统模型。入侵检测核心算法部分,在人工神经网络的基础上,引入了自组织特征映射SOM神经网络。考虑到SOM神经网络连接权向量的随机初始化可能降低入侵检测效率,为此提出优化算法。利用引入模拟退火思想的微粒群算法得到一组全局最优解,将其作为SOM神经网络的初始连接权向量,然后使用SOM神经网络进行模式的训练和分类。搭建Hadoop云平台,部署设计的云环境下入侵检测系统模型。将优化后的入侵检测算法与MapReduce编程模型相结合,实现算法的并行化。通过对比实验验证优化算法能有效提高入侵检测效率,保障云环境安全。
[Abstract]:With the popularization of computer network and the development of information technology, as a new technology, cloud computing is more and more used in various fields of social life. Cloud computing has powerful computing power and storage resources to bring convenience to many enterprises and individuals. At the same time, its huge economic value makes it the target of many hackers. Nowadays, cloud security has become the number one challenge in the development of cloud computing. How to ensure the information security of cloud environment is an urgent task to be solved. Intrusion detection system (IDS) is an important means of dynamic network security protection. Intrusion detection system based on cloud computing has become an important part of cloud security system. Intrusion detection system in cloud environment is also a new development direction in intrusion detection and cloud computing. This paper analyzes the characteristics of intrusion detection technology and cloud computing, combined with the security requirements of cloud environment. Based on CIDF, a model of intrusion detection system in cloud environment is designed. The core algorithm of intrusion detection is based on artificial neural network. The self-organizing feature mapping SOM neural network is introduced. Considering that the random initialization of the connection weight vector of SOM neural network may reduce the efficiency of intrusion detection. For this purpose, an optimization algorithm is proposed. A set of global optimal solutions are obtained by using the particle swarm optimization (PSO) algorithm with the idea of simulated annealing as the initial join weight vector of the SOM neural network. Then the SOM neural network is used to train and classify the patterns. The Hadoop cloud platform is built and the designed intrusion detection system model is deployed in the cloud environment. The optimized intrusion detection algorithm is combined with the MapReduce programming model. The parallel algorithm is realized, and the optimization algorithm is proved to be effective in improving the efficiency of intrusion detection and ensuring the security of cloud environment.
【学位授予单位】：华北电力大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP393.08

【参考文献】