RADIUS: A System for Detecting Anomalous Link Quality Degradation in Wireless Sensor Networks

TL;DR

RADIUS is a lightweight, threshold-based system designed to accurately detect anomalous link quality degradation in wireless sensor networks, maintaining a low error rate even under diverse and dynamic conditions.

Contribution

This paper introduces RADIUS, a novel system that applies Bayes decision theory to improve anomaly detection accuracy in WSNs with minimal error.

Findings

Achieves a stable error rate of 6.13% in indoor tests

Effectively discriminates between good and weak links

Maintains high detection accuracy under dynamic conditions

Abstract

To ensure proper functioning of a Wireless Sensor Network (WSN), it is crucial that the network is able to detect anomalies in communication quality (e.g., RSSI), which may cause performance degradation, so that the network can react accordingly. In this paper, we introduce RADIUS, a lightweight system for the purpose. The design of RADIUS is aimed at minimizing the detection error (caused by normal randomness of RSSI) in discriminating good links from weak links and at reaching high detection accuracy under diverse link conditions and dynamic environment changes. Central to the design is a threshold-based decision approach that has its foundation on the Bayes decision theory. In RADIUS, various techniques are developed to address challenges inherent in applying this approach. In addition, through extensive experiments, proper configuration of the parameters involved in these techniques is identified for an indoor environment. In a prototype implementation of the RADIUS system deployed in an indoor testbed, the results show that RADIUS is accurate in detecting anomalous link quality degradation for all links across the network, maintaining a stable error rate of 6.13% on average.