基于神经网络的指挥自动化网络安全关键技术研究

发布时间：2018-04-24 02:35

本文选题：防火墙 + 入侵检测　；参考：《兰州交通大学》2014年硕士论文

【摘要】：指挥自动化系统作为兵力的倍增器，在复杂多变的战场环境中影响着战斗的进程。指挥自动化网络作为指挥自动化系统功能得以实现的基础，承担着军事信息数据的传输任务，在指挥自动化系统得以发挥战斗力的过程中起着至关重要的作用。指挥自动化网络的安全显得尤为重要，针对当前指挥自动化网络面临的各种威胁，我军采取防火墙、入侵检测等网络安全防御措施。针对传统BP(Back Propagation,反向传播)算法应用于防火墙与入侵检测中，存在收敛速度慢、易陷入局部极小值等缺陷。在传统BP算法基础上进行改进，将小波神经网络以及LM(Levenberg-Marquardt,列文伯格—马夸尔特)算法分别应用于指挥自动化网络防火墙流量的预测和入侵检测的分类中。仿真结果从收敛速度、预测误差、分类效果等方面可以看出，改进算法应用于防火墙流量预测与入侵分类中是十分有效的。具体研究内容如下：首先，分别对防火墙流量的预测以及入侵检测进行建模。根据指挥自动化网络某防火墙流量的特点，研究防火墙流量的影响因素，包括当前防火墙流量数据，星期以及时段等。确定防火墙流量预测输入输出之间的关系，对防火墙流量预测进行建模。入侵检测是防火墙的补充，是一种主动的防御措施。对一些通过伪装等方式进行访问的连接，防火墙无法辨别并进行正确拦截的情况，入侵检测通过主动搜集影响指挥自动化网络关键节点的数据信息，将其作为神经网络的输入。采用KDD CUP99数据集对指挥自动化网络进行建模，得出当前指挥自动化网络是否安全并将入侵进行分类。其次，分别对防火墙流量预测与入侵检测算法进行优化。神经网络算法一个重要的应用领域是预测问题。根据小波算法在时间序列预测问题中效果较好，将其应用于指挥自动化网络防火墙流量的预测中。通过仿真分析，，从收敛速度和预测误差的方面进行对比分析；神经网络算法的另外一个重要的应用领域是解决分类问题。针对传统BP算法分类中存在的缺陷，将LM算法应用于指挥自动化网络入侵分类中。再次，借助Matlab实验平台进行仿真，对参数进行调节，得出最优解。分别对小波与传统BP算法、LM与传统BP算法进行分析、比较。最后，总结指挥自动化网络防火墙流量的预测以及入侵检测中不同算法表现出的优缺点，立足于指挥自动化网络面临的实际情况，提出有待进一步解决的问题。
[Abstract]:As a multiplier of forces, the command automation system affects the process of combat in the complex and changeable battlefield environment. As the basis for the realization of the functions of the command automation system, the command automation network undertakes the task of transmitting military information data, and plays an important role in the process of the command automation system playing a vital role in the process of exerting the combat effectiveness of the command automation system. The security of the command automation network is particularly important. In view of the various threats facing the command automation network, our army adopts network security defense measures such as firewall, intrusion detection and so on. The traditional BP(Back Propagation (back Propagation) algorithm used in firewall and intrusion detection has some shortcomings such as slow convergence rate and easy to fall into local minimum. Based on the traditional BP algorithm, wavelet neural network and Levenberg-Marquardt, Levenberg-Marquardt (Levenberg-Marquardt) algorithm are applied to the traffic prediction and intrusion detection classification of the command automation network firewall, respectively. The simulation results show that the improved algorithm is very effective in firewall traffic prediction and intrusion classification from the aspects of convergence speed, prediction error, classification effect and so on. The specific contents of the study are as follows: Firstly, the prediction of firewall traffic and intrusion detection are modeled. According to the characteristics of firewall traffic in command automation network, the influence factors of firewall traffic are studied, including the current firewall traffic data, week and time period and so on. The relationship between firewall traffic prediction input and output is determined, and firewall traffic prediction is modeled. Intrusion detection is a supplement to firewall and an active defense measure. For some connections which are accessed by camouflage, the firewall can not distinguish and intercept correctly. Intrusion detection takes it as the input of neural network by actively collecting the data information that affects the key nodes of the command automation network. The KDD CUP99 data set is used to model the command automation network, and the security of the current command automation network is obtained and the intrusion is classified. Secondly, the firewall traffic prediction and intrusion detection algorithm are optimized. An important application field of neural network algorithm is the prediction problem. According to the good effect of wavelet algorithm in time series prediction, it is applied to the prediction of firewall traffic in command automation network. Through simulation analysis, the convergence rate and prediction error are compared and analyzed. Another important application field of neural network algorithm is to solve the classification problem. Aiming at the shortcomings of the traditional BP algorithm, LM algorithm is applied to the intrusion classification of command automation network. Thirdly, with the help of Matlab experiment platform, the parameters are adjusted and the optimal solution is obtained. Wavelet and traditional BP algorithm LM and traditional BP algorithm are analyzed and compared. Finally, the paper summarizes the prediction of firewall traffic in command automation network and the advantages and disadvantages of different algorithms in intrusion detection. Based on the actual situation of command automation network, the problems that need to be solved further are put forward.
【学位授予单位】：兰州交通大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TP393.08

【参考文献】