Prevalence of Anomalies in Real World Sensor Network Deployments: The Need for Detection Mechanisms

Prevalence of Anomalies in Real World Sensor Network Deployments: The Need for Detection Mechanisms

Giovani Rimon Abuaitah (Wright State University, USA) and Bin Wang (Wright State University, USA)
Copyright: © 2015 |Pages: 22
DOI: 10.4018/978-1-4666-8251-1.ch006
OnDemand PDF Download:
$37.50

Abstract

Sensor network deployments have enabled automated data collection at a finer granularity compared to human-centric sparse deployments of traditional telemetric data loggers. However, this fine granularity cannot be achieved in the presence of failing nodes, frequent network failures, and/or potentially malicious network behaviors. Moreover, given a perfect and secure network with long-lived sensor nodes, the collected sensor data should be fault-free to guarantee quality of reasoning. Unlike traditional devices, wireless sensor motes are resource-constrained battery-powered devices equipped with few on-board sensors which are susceptible to failures, and batteries cannot always be reliably recharged. Over the years, sensor network deployments have materialized a large number of monitoring and event-detection applications. Researchers and decision makers have observed several anomalous patterns ranging from functional failures to pervasive sensing faults in most of these deployments. In this chapter, we discuss a few of these deployments to emphasize the need for an anomaly detection mechanism.
Chapter Preview
Top

Introduction

Over the last decade, real-life deployments of sensor networks have grown in numbers for various applications. Examples include monitoring weather conditions (Tolle et al., 2005; Szewczyk et al., 2004a), habitat monitoring and wildlife tracking (Szewczyk et al., 2004a; Sorber et al., 2007; Anthony et al., 2012), monitoring of health conditions, flood detection (Ingelrest et al., 2010), fire detection (Liu et al., 2011), volcano monitoring (Werner-Allen et al., 2006), structural health monitoring (Kim et al., 2006), monitoring the micro-climate inside greenhouses (Ahonen et al., 2008), monitoring voltage usages at the micro-grid (Dawson-Haggerty et al., 2012), monitoring water quality (Ramanathan et al., 2006), measuring the quality of perishable food and medicine (Bijwaard et al., 2011), and tracing and verification of goods at different phases of the supply chain (Bijwaard et al., 2011). Most of these deployments have helped decision makers better understand their targeted environments by analyzing the data collected from a large number of unattended sensors. Other deployments have enabled event-driven applications by reporting significant events such as fires in large forests (Liu et al., 2011). A typical deployment of a wireless sensor network (WSN) consists of several tens or hundreds of low-power low-cost tiny devices equipped with sensor boards capable of measuring several phenomena. These devices, usually called sensor motes, are also equipped with wireless communication modules that enable them to communicate with one another. Table 1 shows the characteristics of some sensor platforms (i.e. motes) popularly used in sensor network deployments. Two types of communication paradigms are possible: (a) a simplified single-hop communication where every sensor mote communicates directly with a base station. A base station is a more powerful node which may further process and then relay messages towards a connected PC, and (b) a more complex multi-hop communication paradigm that involves forwarding messages from sensing nodes to the base station with the help of other intermediate sensor nodes in the network (see Figure 1). The base station node is sometimes called the “sink” while other nodes in the network are referred to as “source nodes.”

Complete Chapter List

Search this Book:
Reset