基于多模态数据流的无线传感器网络异常检测方法

发布时间：2018-03-17 16:11

本文选题：无线传感器网络　切入点：时空相关性　出处：《计算机学报》2017年08期 　论文类型：期刊论文

【摘要】：伴随着无线通信技术的不断发展和广泛应用,信息物理融合系统(Cyber-Physical System,CPS)作为物联网领域的最新研究方向成为近年来研究者广泛关注的热点.无线传感器网络(Wireless Sensor Networks,WSN)作为CPS系统物理空间的主要感知网络,若有效提高对应感知数据的准确性和可靠性,可及时准确地发现突发事件、监测网络工作状况,因此对传感器网络节点数据流进行异常检测,发现其中的异常数据并确认其来源具有重要意义.该文在无线传感器网络多模态数据流研究的基础之上,提出了一种对传感器异常数据进行检测以及监测节点自身工作状态的方法,该方法不仅应用了无线传感器网络中的时空相关性原理,还更进一步,研究了同一节点中多模态数据流之间的相干性,并以此作为理论基础,利用多维数据和滑动窗口模型对异常数据及其来源进行检测和验证.该文的方法可以分为3个步骤:首先,利用滑动窗口中的历史数据对传感器数据流进行异常数据的检测;其次,利用节点的空间相关性对异常的来源进行确认和识别;最后,对由于测量误差导致的异常值进行筛选,使输入CPS的数据进一步的精确化.通过实验对比,该文的方法对传感器异常数据的检测率保持在95%;在不同数据维度的条件下,对四维数据集的检测率比单维数据集提高了3%.
[Abstract]:With the continuous development and wide application of wireless communication technology, As the newest research direction in the field of Internet of things, the information physics fusion system (Cyber-physical system CPS) has become a hot topic in recent years. Wireless Sensor Networks (WSNs) is the main perceptual network in the physical space of CPS system. If the accuracy and reliability of the corresponding perceptual data are improved effectively, the unexpected events can be detected in time and accurately, and the network working condition can be monitored, so the abnormal detection of the sensor network node data flow can be carried out. It is of great significance to discover the abnormal data and confirm its source. Based on the research of multi-modal data flow in wireless sensor networks, This paper presents a method to detect abnormal data of sensor and monitor the working state of nodes. This method not only applies the principle of space-time correlation in wireless sensor networks, but also goes further. The coherence between multimodal data streams in the same node is studied and used as the theoretical basis. Using multi-dimension data and sliding window model to detect and verify the abnormal data and its source, the method can be divided into three steps: firstly, using the historical data in the sliding window to detect the abnormal data of sensor data flow; Secondly, the source of the anomaly is confirmed and identified by the spatial correlation of the node. Finally, the outlier value caused by the measurement error is screened to make the input CPS data more accurate. In this paper, the detection rate of the sensor anomaly data is kept at 95, and the detection rate of the four-dimensional data set is increased by 3 times than that of the single-dimensional data set under different data dimensions.
【作者单位】：南京邮电大学计算机学院;江苏省无线传感网高技术研究重点实验室;江南大学物联网工程学院;
【基金】：国家自然科学基金(61472368,61373137,61572260) 江苏省高校自然科学(14KJA520002) 江苏省六大人才高峰项目基金(2013-DZXX-014)资助~~
【分类号】：TN929.5;TP212.9

【参考文献】