摘要
在入侵检测中,传统的极限学习机(ELM)没有考虑到两方面的问题:一是误分类代价的敏感问题,在网络入侵检测中,需要考虑到误分类代价不同的问题,并以实现期望代价最小化为目标;二是冗余特征的处理问题,当入侵特征维数较多时,会存在着大量的冗余特征的问题,如果直接对高维数据进行分类,不仅入侵模式不能被准确分类,误检率较高,并且大量的冗余数据既耗费了系统的资源,也增大了入侵检测的时间。针对这两方面的问题,提出一种基于主成分分析法的代价敏感极限学习机(ELM)。通过主成分分析法对数据进行降维,确定主要特征;再将预处理后的数据训练极限学习机(ELM),以实现期望代价最小化为目标,从而实现降低入侵检测的检测时间,降低检测误报率,提高检测准确率的目的。实验表明,在入侵检测中,基于主成分析法的代价敏感极限学习机(ELM)与传统的ELM相比,不仅使分类准确率得到提高,降低了分类的误报率,而且在分类速度上也有一定的优越性,提高了网络运行的效率。
In the intrusion detection system,the traditional ELM does not consider the two aspects.One is the sensitive problem of the cost of misclassification.In network intrusion detection,it is necessary to consider the different costs of misclassification,and to minimize the expected cost.The other is the processing problem of redundant features.The more dimensions of the intrusion features,the more redundant features.If the high-dimensional data are classified directly,not only the intrusion modes can not be accurately classified,the false detection rate is high,and a large number of redundant data will consume the resources of the system,and increase the time of intrusion detection.As for these two problems,a cost sensitive ELM based on principal component analysis is proposed.It reduces the data through the main analysis method,and determines the mainfeatures,and then quickly classifies the trained data through cost sensitive ELM,to minimize the expected cost,so as to reduce the detection time of intrusion detection,reduce the detection error rate,and improve detection accuracy.The experiment shows that in the intrusion detection,the cost sensitive ELM based on the principal component analysis method,compared with the traditional ELM,not only improves the classification accuracy,but also has a certain superiority in the classification speed,and improves the efficiency of the network operation.
作者
顾竞豪
王晓丹
GU Jing-hao;WANG Xiao-dan(Institute of Air Defense and Anti-Missile,Air Force Engineering University,Xi’an 710051,China)
出处
《火力与指挥控制》
CSCD
北大核心
2020年第1期139-143,共5页
Fire Control & Command Control
关键词
主成分分析法
代价敏感
极限学习机
数据处理
principal component analysis
cost sensitive
extreme learing machine
data processing
