网络入侵检测是一种基于网络行为特征的检测技术.近年来,作为信息安全领域中的研究热点,网络入侵检测发展迅速.针对传统入侵检测算法对于数据特征提取较慢的问题,本文提出了基于信息熵理论的免疫算法来提高特征提取速度.为了进一步提高...网络入侵检测是一种基于网络行为特征的检测技术.近年来,作为信息安全领域中的研究热点,网络入侵检测发展迅速.针对传统入侵检测算法对于数据特征提取较慢的问题,本文提出了基于信息熵理论的免疫算法来提高特征提取速度.为了进一步提高分类精度,本文对Adaboost分类方法进行了改进,在分类过程中判断噪声数据,并对噪声数据的权重进行调整,从而缓解了Adaboost算法的过度拟合.通过对KDD CUP 99数据的实验结果表明,本文方法可以提高免疫算法在特征提取方面的收敛速度,并能有效地提高入侵检测率.展开更多
A particle swarm optimization (PSO) algorithm improved by immunity algorithm (IA) was presented. Memory and self-regulation mechanisms of IA were used to avoid PSO plunging into local optima. Vaccination and immune se...A particle swarm optimization (PSO) algorithm improved by immunity algorithm (IA) was presented. Memory and self-regulation mechanisms of IA were used to avoid PSO plunging into local optima. Vaccination and immune selection mechanisms were used to prevent the undulate phenomenon during the evolutionary process. The algorithm was introduced through an application in the direct maintenance cost (DMC) estimation of aircraft components. Experiments results show that the algorithm can compute simply and run quickly. It resolves the combinatorial optimization problem of component DMC estimation with simple and available parameters. And it has higher accuracy than individual methods, such as PLS, BP and v-SVM, and also has better performance than other combined methods, such as basic PSO and BP neural network.展开更多
文摘网络入侵检测是一种基于网络行为特征的检测技术.近年来,作为信息安全领域中的研究热点,网络入侵检测发展迅速.针对传统入侵检测算法对于数据特征提取较慢的问题,本文提出了基于信息熵理论的免疫算法来提高特征提取速度.为了进一步提高分类精度,本文对Adaboost分类方法进行了改进,在分类过程中判断噪声数据,并对噪声数据的权重进行调整,从而缓解了Adaboost算法的过度拟合.通过对KDD CUP 99数据的实验结果表明,本文方法可以提高免疫算法在特征提取方面的收敛速度,并能有效地提高入侵检测率.
文摘A particle swarm optimization (PSO) algorithm improved by immunity algorithm (IA) was presented. Memory and self-regulation mechanisms of IA were used to avoid PSO plunging into local optima. Vaccination and immune selection mechanisms were used to prevent the undulate phenomenon during the evolutionary process. The algorithm was introduced through an application in the direct maintenance cost (DMC) estimation of aircraft components. Experiments results show that the algorithm can compute simply and run quickly. It resolves the combinatorial optimization problem of component DMC estimation with simple and available parameters. And it has higher accuracy than individual methods, such as PLS, BP and v-SVM, and also has better performance than other combined methods, such as basic PSO and BP neural network.
