基于神经网络的入侵检测技术的研究.docxVIP

摘 要 因特网的出现正影响着人类的生存方式，推动了世界的发展进程。网络安全变得越来越重要，如何保护网络避免受到网络攻击迫在眉睫。网络安全的主要目的是监视和保持传入和传出流量的检查，以防止任何恶意活动和策略违规。入侵检测技术作为一种积极主动的动态安全技术，提供了对内部攻击、外部攻击和误操作的实时保护。入侵检测是保护网络信息数据的有效工具，成为信息安全领域的研究热点。 本文通过分析传统的BP神经网络的入侵检测技术有一定的缺点，于是提出了采用粒子群优化算法来改进BP神经网络的方法，实现了提高识别率、可读性以及执行速度的目的。对绝大多数是多维和嘈杂的数据，进行了数据降维，通过主成分分析方法处理多维数据。在本文神经网络的实验中，通过比较PSO-BP（Particle Swarm Optimization-Back Propagation）和PCA-PSO-BP（Principal Component Analysis-Particle Swarm Optimization-Back Propagation）两种算法的比较得出PSO-BP相较于BP（Back Propagation），PCA-PSO-BP相较于PSO-BP各方面都有了改善。主成分分析特征降维方法以及优化后的PSO-BP算法都能够降低网络训练的误检率和漏报率，缩短训练收敛速度和检测时间，提高了预测准确率。 关键词：网络安全；入侵检测；BP神经网络；特征降维 ABSTRACT The emergence of the Internet is affecting the way of human growth, and promoting the development process of the world. Network security has become increasingly important. It’s imminent to protect the network to avoid being attacked by the hackers. The primary purpose of network security is to monitor and maintain incoming and outgoing traffic checks to prevent any malicious activity and policy violations. Intrusion detection technology as a proactive dynamic security technology, provides internal attacks, external attacks and misuse of real-time protection. Intrusion detection is an effective tool to protect network information and become a hotspot in information security field. This paper analyzes the traditional BP neural network intrusion and detection technology. It also has some shortcomings. We put forward the use of particle swarm optimization algorithm to improve the BP neural network method to achieve the purpose of improving the recognition rate, readability and execution speed. For the vast majority of multidimensional and noisy data, the data dimensionality is reduced. Multidimensional data is processed by principal component analysis. In the experiment of neural network in this paper, PSO-BP was compared with PCA-PSO-BP. PCA-PSO-BP is better. The principal component analysis features reduction method and the optimized PSO-BP algorithm can