基于云框架直方图的无线传感网络数据异常检测方法

Data Anomaly Detection Method for Wireless Sensor Networks Based on Cloud Framework Histogram

以提高无线传感网络环境安全性为目的,提出一种基于云框架直方图的数据异常检测方法。在高斯函数分类云模型中,划分不同数据的属性中心值,完成数据高维分类。然后针对可能存在异常数据的传感网络环境,在构建云框架直方图后计算数据间相似度。通过比较低的数据间相似度阈值完成数据异常粗检测,再在不同维度空间内,利用超矩形方法对比粗检测后数据是否处于同一数段,实现数据异常细检测。仿真结果表明:所提方法的归一化训练结果基本符合归一化处理的理论变化趋势,在数据量最大的情况下,其检测耗时仅用41 s、检测误差仅为1.2%,且在中心机房的能耗中占比较小。相比于传统方法来说,所提方法的检测耗时更少、误差更小、检测能耗更低。

In order to improve the security of wireless sensor network environment,a data anomaly detection method based on cloud framework histogram is proposed.In the classification cloud model using Gaussian function,the attribute center values of different data are divided to complete high-dimensional classification of the data.Then,aiming at the sensor network environment where abnormal data may exist,the similarity among data is calculated after constructing the cloud framework histogram.The anomaly coarse detection of the data is completed through the threshold of low similarity,and then the hyperrectangle method is used to compare whether the data after coarse detection is in the same number of segments in different dimensional spaces to achieve anomaly fine detection of the data.The simulation results show that the normalized training results of the proposed method basically conform to the theoretical trend of normalized processing.In the case of the maximum amount of the data,the detection time is only 41 seconds,the detection error is only 1.2%,and the energy consumption of the central machine room is relatively low.Compared with the traditional method,the detection time of the proposed method is less,the error is less,and the detection energy consumption is lower.