基于MSNN模型的网络安全入侵检测
2019-08-13朱韶平肖永良党艳军
朱韶平 肖永良 党艳军



摘 要:为解决网络系统入侵行为升级快,隐蔽性强和随机性高等严重安全问题,结合入侵检测系统信息的特点,提出一种基于MSNN模型的入侵检测算法。首先提取系统调用顺序特性和频度特性,然后利用多级Sigmoid神经网络中的Sigmoid神经元具有微调网络的作用,且能让神经元产生多元反应进行多类分类,构建类似于大脑神经突触网络信息处理的MSNN模型,实现网络安全入侵检测。实验结果表明,该算法的检测精度高、抗干扰能力强,具有良好的检测效果和较高的应用价值。
关键词:网络安全;入侵检测;MSNN模型;系统调用顺序特性;系统调用频度特性
中图分类号:TP391.4 文献标识码:A
Intrusion Detection of Network Security Based on MSNN Model
ZHU Shao-ping1?覮,XIAO Yong-lian2,DANG Yan-jun1
(1. Department of Electronic Information Engineering,Zhuhai City Polytechnic,Zhuhai,Guangdong 519090,China;
2. Department of Information Management,Hunan University of Finance and Economics,Changsha,Hunan 410205,China)
Abstract:In order to solve the serious security problems of network system intrusion behavior,such as rapid upgrade,strong concealment and high randomness,an intrusion detection algorithm based on MSNN model is proposed in combination with the characteristics of intrusion detection system information. First extract the system transfer sequence characteristics and frequency characteristics,then the algorithm use Sigmoid neurons in the multilevel Sigmoid neural network to fine-tune the network and enable the neurons to generate multiple responses for multiple classification,so as to build an MSNN model similar to the brain's synaptic network information processing and realize network security intrusion detection. The experimental results show that the proposed algorithm has high precision and strong anti-interference ability,and has a good detection effect and high application.
Key words:network security;intrusion detection;MSNN model;order characters of system;frequency characters of system
隨着物联网、移动互联网的技术的飞速发展,网络攻击逐渐呈现智能化和复杂化的趋势,人们面临的网络安全问题日益严峻。入侵检测技术是通过分析从计算机系统或计算机网络中收集来的安全日志和网络数据包等信息,检测出计算机系统或计算机网络中可能存在的违反安全策略的入侵行
为 [1]。它是一种重要网络安全防护手段,能为网络用户提供隐私和数据保护,是近年来信息安全领域的研究热点。早在1987年Dorothy Denning[2]提出了第一个入侵检测模型;1998年,IDES和Chen[3]等人相继提出了通用入侵检测模型CIDF;Ye N,Li X[4] 等人应用概率理论提出了基于计算机审计数据的入侵检测模型,并取得了较好的检测效果;张玲等人综合误用检测和异常检测提出了一种基于粗糙集和人工免疫的集成入侵检测(RSAI-IID)模型[5];张得生等人结合SVM 和融合技术提出了基于SVM和融合技术的入侵检测,有效解决了传统SVM 算法易产生训练参数选择不当、分类精度低等问题[6]。……
