Abstract: Based on the study of machine learning algorithm,this paper proposes a hybrid intrusion detection scheme using the Extreme Learning Machine(ELM)for Wireless Sensor Network(WSN). It divides the WSN into the perception layer,data aggregation layer and the core control layer,corresponding intrusion detection scheme is presented at each layer.Especially,in the core control layer,it sets trust management modules and ELM modules. Using a trust module can timely sieve to abnormal nodes. The ELM is faster than the SVM algorithm,and the efficiency of intrusion detection can be further improved when using ELM. Experimental results show that the scheme that combines the ELM with traditional intrusion detection technology balances the advantages and disadvantages,reduces the energy consumption and prolongs the network uptime,on the basis of guaranteeing a higher detection rate. So it is more suitable for WSN which is resource-constrained.

Key words: Wireless Sensor Network(WSN), Extreme Learning Machine(ELM), hybrid, intrusion detection, trust management, clustering

摘要： 在研究机器学习算法的基础上,提出一种基于极限学习机(ELM)的混合入侵检测方案。将无线传感器网络分为感知层、数据汇聚层和核心控制层,在每层分别设置与其相适应的入侵检测方案,并在能量充足的核心控制层布置信任管理模块和ELM 模块。信任模块可以及时筛去异常节点,相比于支持向量机算法训练速度更快,可提高入侵检测效率。实验结果表明,该方案在保证较高检测率的基础上,降低了能耗,延长网络运行时间,更适合于资源受限的无线传感器网络。

关键词: 无线传感器网络, 极限学习机, 混合, 入侵检测, 信任管理, 分簇

CLC Number: 

• TP309