摘要
入侵检测系统中的特征选择是一个组合优化问题。为了有效地进行特征选择,提出一种结合进化思想的免疫算法。算法中的免疫记忆单元确保了快速收敛于全局最优解,算法中的均匀交叉操作则体现了进化的思想。提出一个基于神经网络的入侵检测系统模型,该模型具有多分类,易于更新系统使其快速适应新型入侵的特点。在KDD CUP’99上的实验表明该算法是有效的。
Feature selection in intrusion detection system is an optimization problem. An immune algorithm combined evolutional spirit is proposed in this paper in order to select features effectively. The immune memory units guarantee this algorithm rapid convergence to global optimum and the uniform crossover operator embody the idea of evolution. Furthermore, a model of intrusion detection system based on Neural Networks is presented. The model characterizes itself in muticlassifications and updating easily to adapt new intrusion modes. Experiments on KDD CUP'99 indicate the effectiveness of this algorithm presented in this paper.
出处
《微电子学与计算机》
CSCD
北大核心
2007年第3期20-22,26,共4页
Microelectronics & Computer
基金
国家自然科学基金项目(60475007)
教育部跨世纪人才基金项目
关键词
入侵检测系统
免疫算法
记忆单元
神经网络
intrusion detection system
immune algorithm
memory unit
neural networks