基于四分之一超球SVM的WSN异常检测

Anomaly detection in WSNs based on quarter-sphere SVM

无线传感器网络(WSN)收集的数据本质上是不可靠的,因此为了提高数据质量,需要对网络进行异常值检测。文中提出了一种基于四分之一超球支持向量机(SVM)算法的异常数据检测方法,利用从传感器节点中收集到的原始数据建立支持向量机预测模型,并结合粒子群算法(PSO)找出最佳参数,然后利用最佳参数对原本的模型进行优化。以一种分布式在线方式,对正常和异常数据进行实时区分。实验结果表明,该方法可以实现异常检测的效果,并且具有较高的准确率和较低的误报率。

Data collected by wireless sensor networks( WSN) are essentially unreliable.To improve the quality of data,an anomaly detection of the network is needed.An anomaly data detection method based on quarter-sphere support vector machine( SVM) is proposed.Firstly, the SVC prediction model is established by using the raw data collected from the sensor nodes.Then,by combined with particle swarm optimization ( PSO), the optimal parameters are found.Finally, the differentiation of normal and abnormal data is implemented in a distributed and online manner.Experimental results show that the anomaly detection technique can detect outliers,and has higher detection accuracy and lower false positive rate.