摘要
针对入侵检测的特征和分类器参数选择问题,采用极限学习机ELM(extreme learning machine)进行构建分类器,提出一种蝙蝠算法(BA)联合选择特征和分类器参数的网络入侵检测模型(BA-ELM)。首先将特征子集和极限学习机参数编码成蝙蝠个体,以入侵检测准确率和特征数加权组成个体适应度函数;然后通过个体和群体更新的规则引导蝙蝠向最优解飞行,从而找到最优的子特征集和极限学习机参数;最后建立最优的入侵检测模型,并通KDD CUP 99数据集进行仿真性能分析。结果表明,BA-ELM较好地解决了入侵检测特征选择与分类器参数不匹配难题,提高了网络入侵检测率和检测效率,更加适合于网络入侵检测的实时要求。
Aiming at the problems of intrusion detection features and classifier parameter selection, we use extreme learning machine （ELM） to construct classifier, and propose an intrusion detection model （BA-ELM） which is based on jointly selecting features and classifier parameters by bat algorithm （BA）. First, we encode the feature subset and the ELM parameters to bat individuals, and compose the intrusion detection accuracy and the feature number weighting to individual fitness function; then we guide the bats flying toward the optimal solution through individual and population update rules so that to find optimal feature subset and ELM parameters; finally, we build optimal intrusion detection model, and analyse simulation performance through KDD Cup 99 dataset. Results show that the BA-ELM well solves the problem of mismatching between intrusion detection feature selection and classifier parameter and improves the network intrusion detection rate and detectionefficiency, it is more suitable for the real-time requirements of network intrusion detection.
出处
《计算机应用与软件》
CSCD
北大核心
2014年第7期294-296,306,共4页
Computer Applications and Software
关键词
特征选择
分类器参数
极限学习机
蝙蝠算法
Feature selection
Classifier parameters
Extreme learning machine
Bat algorithm
