Malicious and Malfunctioning Node Detection via Observed Physical Layer Data

摘要

There are many mechanisms that can cause inadequate or unreliable information in sensor networks. A user of the network might be interested in detecting and classifying specific sensors nodes causing these problems. Several network layer based trust methods have been developed in previous research to assess these issues; in contrast this work develops a trust protocol based on observations of physical layer data collected by the sensors. Observations of physical layer data are used for decisions and calculations, and are based on just the measurements collected by the sensors. Although this information is packaged and distributed on the network layer, only the physical measurement is considered. This protocol is used to detect faulty nodes operating in the sensor network. The context of this research is Wireless Network Discovery (WND), which refers to modeling all layers of a non- cooperative wireless network. The focus in particular is the localization of transmitters, and detection of sensors affecting the localization. To accomplish this, a model for faulty sensors and two methods of detection are developed. Detection rates are analyzed with Receiver Operating Characteristic (ROC) curves, and the trade-off of detection versus localization error is discussed. Classification between faulty sensors is also considered to determine appropriate response to potential network attacks.