项目名称: 无线传感器网络网内异常检测技术研究
项目编号: No.61202046
项目类型: 青年科学基金项目
立项/批准年度: 2013
项目学科: 计算机科学学科
项目作者: 吴中博
作者单位: 湖北文理学院
项目金额: 23万元
中文摘要: 传感器网络目前被广泛应用于国防军事、环境监测、交通管理、智能家居、制造业、反恐抗灾等领域。复杂的环境和低廉的价格导致传感器节点易于失效,失效的节点可能会报告任意的读数;而且,由于受到干扰,传感器节点也经常会产生噪音。因此,异常检测是传感器网络走向实用的重要支撑技术。现有的算法大多考虑减少通讯开销,忽略了节点的实际计算能力;而且现有的异常检测和查询处理是两个孤立的过程,不能很好结合。本项目拟提出一种近似的异常检测技术,在簇头和簇间建立两级异常检测机制,可以有效平衡网络负载;利用节点向量相似性来进行异常检测,针对节点相似性比较时间复杂度过高的问题,利用局部敏感哈希映射机制对节点读向量进行缩减,在保证相似性比较精度的同时减少时间复杂度;提出一种面向查询处理的异常检测技术,在查询处理的同时进行异常检测,提高查询处理结果的可靠性,进一步促进和推动无线传感器网络实用化的进程。
中文关键词: 传感器网络;异常检测;网内处理;相似性;
英文摘要: Wireless Sensor Network has be widely used in the fields of national defense and military, environmental monitoring, city management, smart home, manufacturing industry, anti-terrorism and disaster resistant etc. Sensor nodes are easy to be fail because of complex environment and cheap price. Fail sensor nodes may send random data. Furthermore, sensor nodes may produce noise data because of interference. So outlier detection is an important technology for the practicality of sensor network. Existing algorithms focus on reducing communication cost and omit sensor nodes' computation ability. Furthermore, existing outlier detection and query processing are two isolated process and they can not combine with each other directly. In this research we will put forward an approximate outlier detection algorithm which builds two-level detection mechanism in both cluster header and adjacent clusters using node vector similarity to detect outlier to balance network overload. According to the Complexity of vector similarity comparison, we use Locality Sensitive Hashing method to reduce node vector which can ensure comparison precision and reduce computational complexity. We will also put forward an outlier detection method oriented to query processing which detects outlier and executes query at the same time to improve the
英文关键词: Sensor Network;Outlier Detection;In-Network process;Similarity;