基于约简策略与自适应SVDD的无线传感网络离群检测方法

发布时间：2018-06-04 11:38

本文选题：无线传感网络 + 数据约简　；参考：《传感技术学报》2017年09期

【摘要】：无线传感网络经常被部署于条件恶劣、无人值守的环境中,受到恶劣天气、软硬件故障、能量不足或者恶意攻击等因素的影响,传感器节点感知数据的缺失或错误难以避免。因此,传感器数据流的离群检测对于提高系统可用性至关重要。提出一种基于约简策略与自适应SVDD(Support Vector Data Description)的离群检测方法(RASVDD),该方法首先使用基于马氏距离标准的方法约简数据集,有效地减少了训练样本;然后利用基于数据分布密度准则和数据流时间相关性自适应更新决策模型。针对Intel Berkeley数据集及Sensor Scope System数据集的仿真实验表明,RASVDD的离群检测正确率TPR(True Positive Rate)平均达98%,误报率FPR(False Positive Rate)平均仅为1%,并且与传统的SVDD算法相比,RASVDD决策模型训练时间减少了20%以上。
[Abstract]:Wireless sensor networks are often deployed in harsh and unattended environments, which are affected by bad weather, hardware and software failures, insufficient energy or malicious attacks, and the loss or error of sensor nodes' perceptual data is difficult to avoid. Therefore, outlier detection of sensor data stream is very important to improve system availability. A method of outlier detection based on reduction strategy and adaptive SVDD(Support Vector Data Description) is proposed. Firstly, the method based on Markov distance criterion is used to reduce the data set, which effectively reduces the training samples. Then the decision model is updated adaptively based on data distribution density criterion and data stream time correlation. The simulation results on Intel Berkeley data sets and Sensor Scope System datasets show that the correct rate of outlier detection is 98 and the FPR(False Positive rate of false alarm is only 1. Compared with the traditional SVDD algorithm, the training time of RASVDD decision model is reduced by more than 20%.
【作者单位】：江南大学物联网工程学院;物联网技术应用教育部工程技术研究中心;江苏省无线传感网高技术研究重点实验室;
【基金】：国家自然科学基金项目(61472368,61174023) 江苏省重点研发计划项目(BE2016627)
【分类号】：TN929.5;TP212.9

【相似文献】