摘要
高维网络数据中的无关属性和冗余属性容易使分类算法的网络入侵检测速度变慢、检测率降低。为此,提出一种基于遗传量子粒子群优化(GQPSO)算法的网络入侵特征选择方法,该方法将遗传算法中的选择变异策略与QPSO有机结合形成GQPSO算法,并以网络数据属性之间的归一化互信息量作为该算法适应度函数,指导其对网络数据的属性约简,实现网络入侵特征子集的优化选择。在KDDCUP1999数据集上进行仿真实验,结果表明,与QPSO算法、PSO算法相比,该方法能更有效地精简网络数据特征,提高分类算法的网络入侵检测速度及检测率。
Aiming at problem that independent and redundant attributes of high dimensional network data cause classification algorithms＇ slow detection speed and low detection rate in network intrusion detection, a feature selection approach for network intrusion based on Genetic Quantum Particle Swarm Optimization（GQPSO） algorithm is proposed. The approach organically combines selection and variation of genetic algorithm with QPSO to form GQPSO algorithm, and normalizes mutual information between attributes of network data is defined as the algorithm＇s fitness function, which guides its reduction of network data attributes to realize optimal selection of network intrusion feature sub-set. Simulation experiment is done in KDDCUP1999. Result shows that compared with QPSO and PSO algorithms, the approach is more effective for feature selection of network data and improvement of network intrusion detection speed and detection rate of classification algorithms.
出处
《计算机工程》
CAS
CSCD
北大核心
2011年第14期103-105,共3页
Computer Engineering
基金
陕西省自然科学基金资助项目(2009JM7007)
关键词
GQPSO算法
归一化互信息
适应度函数
特征选择
网络入侵检测
Genetic Quantum Particle Swarm Optimization（GQPSO） algorithm
normalized mutual information
fitness function
featureselection
network intrusion detection
