基于S-Kohonen神经网络的拒绝服务攻击检测算法的研究

发布时间：2018-01-23 21:19

本文关键词： 信息安全拒绝服务 S-Kohonen神经网络学习向量量化入侵检测　出处：《河北大学》2017年硕士论文　论文类型：学位论文

【摘要】：拒绝服务攻击(Denial of Service,DoS)是一种既简单又有效的网络攻击方式。它企图使其预期用户的一台主机或其他网络资源不可用,如临时或无限期地中断或暂停连接到因特网主机的服务。随着人们越来越依赖于网络,因而导致潜在的受害者越来越多。本文研究了基于Kohonen神经网络的DoS攻击检测算法。现有的DoS攻击检测方法中,基于人工神经网络的算法是较为有效的一类方法,其优点是并行处理能力强,对噪声数据有较强的鲁棒性和容错能力,能充分逼近复杂的非线性关系,具备联想记忆功能等。但此类算法也存在着收敛速度慢和不能使同类神经元逐渐集中等的问题,为克服上述问题,本文做了如下研究工作:(1)通过调整Kohonen网络的学习速率和学习半径,并增加输出层使其转化为有监督的学习网络,从而有效缓解了Kohonen网络收敛速度慢的问题。(2)通过增加一层网络权值,并对网络权值进行调整,最终使得获胜神经元及周边神经元权值更加接近该输入样本,从而在一定程度上克服了同类神经元不能较快集中的问题,并提出了基于S-Kohonen的拒绝服务攻击检测算法。(3)为了使基于S-Kohonen的拒绝服务攻击检测算法更完善,在竞争层上不能自动学习输入向量模式的分类问题上,将LVQ算法与S-Kohonen网络相结合,使其可以对输入向量模式的分类自主学习,降低网络结构,提高算法的检测率和减少运算时间。本文提出了一种基于S-Kohonen的DoS攻击检测算法,将改进的Kohonen网络引入到了DoS攻击检测算法中;并结合LVQ算法对本文所提出的算法进一步进行优化,仿真实验表明,所提算法有效。
[Abstract]:A denial of service attack (Denial of, Service, DoS) is a simple and effective way of network attacks. It attempts to make its intended users a host or other cyber source is not available, such as temporary or indefinite interruption or host connected to the Internet service is suspended. As people increasingly dependent on the network, due to the growing number of victims of potential. This paper studies the DoS attack detection algorithm based on Kohonen neural network. DoS attack detection method in the existing algorithm based on artificial neural network is a kind of effective method, the utility model has the advantages of parallel processing capability, robustness and strong fault tolerance to noise data. Can fully approach the complex nonlinear relationship, have the function of associative memory. But this algorithm also has slow convergence speed and can not make the same neuron gradually concentrated and other issues, to overcome the problems on the, This research work done as follows: (1) the learning rate and the radius of learning adjustment of Kohonen network, and increase the output layer transformed into supervised learning network, so as to effectively alleviate the problem of slow convergence speed of the Kohonen network. (2) by adding a layer of network weights, and adjust the weights of the network. The winning neuron and the surrounding neuron weights closer to the input sample, thus the same neuron cannot quickly focus on the problem to overcome to a certain extent, and proposed a denial of service attack detection algorithm based on S-Kohonen. (3) in order to make a denial of service attack detection algorithm based on S-Kohonen is more perfect, in the competition layer can automatically learn the input vector pattern classification problems, combines LVQ algorithm and S-Kohonen network classification, so that it can be the input vector model of autonomous learning, reduce network structure, improve the Algorithm of detection rate and reduce the computation time. This paper presents a DoS based S-Kohonen attack detection algorithm, the improved Kohonen network is introduced to the DoS attack detection algorithm; and LVQ algorithm of the proposed algorithm is further optimized, simulation results show that the proposed algorithm is effective.

【学位授予单位】：河北大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP183;TP393.08

【相似文献】