摘要
针对无线传感器网络在定位过程中的外部攻击节点的类型识别问题,提出了一种交替方向-Lp范数支持向量机(ADM-PSVM)分布式识别算法。该算法基于线性支持向量机分类模型,首先引入了Lp范数约束形式,通过选择不同的范数值p以增强分类算法对数据集的适应能力;继而根据交替方向乘子方法推导出了算法的分布式形式,实现了节点根据剩余能量将识别的计算任务分布于不同节点之间进行;最后将算法对各类型的恶意节点数据进行了训练及识别仿真,并讨论了范数约束值以及惩罚因子取值的不同对识别精确率的影响。仿真结果表明,该算法对于恶意外部攻击节点数据具有较好的识别精确度及更高的计算效率。
The process of localization in wireless sensor networks is easily attacked by malicious nodes. In order to identify the types of those external attacks, an Alternating Direction Method of Multipliers-p-Norm Support Vector Machines (ADM-PSVM) algorithm is proposed. The proposed algorithm is based on classification model of the linear support vector machine. Firstly, by introducing a norm constraint into the classification algorithm, the adaptability of classifier for various types of dataset can be enhanced via selecting different value p. Then we derive distributed form of the algorithm according to Alternating Direction Method of Multipliers; this makes the classifier have the ability to distribute computing task among different nodes based on the residual energy of each node. Finally, the sample and testing dataset for each of four types of external malicious nodes are implemented in the training and tes- ting processes of the proposed algorithm, and the influence on recognition accuracy performance in various p values and penalty factor η ones are discussed. The experimental results show that the proposed algorithm can achieve higher classification accuracy and better computational efficiency on the hostile external attack dataset.
出处
《西北工业大学学报》
EI
CAS
CSCD
北大核心
2016年第1期85-91,共7页
Journal of Northwestern Polytechnical University
基金
国家自然科学基金(61401360)
陕西省自然科学基础研究计划(2014JQ2-6033)
中央高校基本科研业务费专项资金(3102014JCQ01055)资助
关键词
分布式
支持向量机
传感器网络
p范数
定位
识别
adaptive systems, classifiers, computational efficiency, eigenvalues and eigenfunctions, iterative meth-ods, Lagrange multipliers, matrix algebra, mesh generation,sampling,support vector machines, vec-tots, wireless sensor networks
ADM-PSVM (Alternating Direction Method of Multipliers-p-Norm Sup-port Vector Machines), attack type recognition, classification, distributed, localization, malicious,p-norm