摘要
为了在有限算法复杂度的基础上提高无线传感器网络的攻击检测率,提出了一种改进的支持向量机多类分类算法.该算法综合了稀疏型随机编码和Hadamard编码的特点,以汉明距离为评判依据,对节点采集的流量数据进行分类.结果表明,与单独的一对一、一对多及Hadamard算法相比,此改进型分类算法在五种攻击的正确率检测方面有较明显的优势,运算时间上比Hadamard算法减少了22%.
In order to improve the attack detection rate in wireless sensor networks on the basis of limited algorithm complexity,a modified multi-class classification algorithm of support vector machine(SVM)was proposed.The algorithm which integrated the characteristics of sparse random coding and Hadamard coding was used to classify the traffic data collected by sensor nodes according to the judging criteria of hamming distance.The results showed that the modified classification algorithm is more excellent in detection accuracy of five kinds of attack compared with one-against-one,one-against-all and Hadamard method.Besides,the operation time of the modified algorithm is 22percent less than that of Hadamard algorithm.
出处
《中国计量学院学报》
2013年第3期298-303,共6页
Journal of China Jiliang University
基金
国家自然科学基金资助项目(No.61027005/F010906)
关键词
无线传感网络
支持向量机
入侵检测
wireless sensor networks
support vector machine
intrusion detection