Learning to Detect Anomalous Wireless Links in IoT Networks

TL;DR

This paper investigates automatic anomaly detection in IoT wireless links using machine learning, demonstrating high accuracy with supervised and unsupervised methods, especially OC-SVM, in a real-world deployment.

Contribution

It introduces four types of wireless link anomalies in IoT and evaluates ML classifiers, highlighting the effectiveness of supervised and unsupervised approaches, particularly OC-SVM.

Findings

Supervised ML approaches achieve F1 scores above 0.98.

Unsupervised ML approaches reach F1 scores around 0.90.

OC-SVM outperforms other unsupervised methods with F1 scores up to 0.99.

Abstract

After decades of research, the Internet of Things (IoT) is finally permeating real-life and helps improve the efficiency of infrastructures and processes as well as our health. As a massive number of IoT devices are deployed, they naturally incur great operational costs to ensure intended operations. To effectively handle such intended operations in massive IoT networks, automatic detection of malfunctioning, namely anomaly detection, becomes a critical but challenging task. In this paper, motivated by a real-world experimental IoT deployment, we introduce four types of wireless network anomalies that are identified at the link layer. We study the performance of threshold- and machine learning (ML)-based classifiers to automatically detect these anomalies. We examine the relative performance of three supervised and three unsupervised ML techniques on both non-encoded and encoded (autoencoder) feature representations. Our results demonstrate that; i) selected supervised approaches are able to detect anomalies with F1 scores of above 0.98, while unsupervised ones are also capable of detecting the said anomalies with F1 scores of, on average, 0.90, and ii) OC-SVM outperforms all the other unsupervised ML approaches reaching at F1 scores of 0.99 for SuddenD, 0.95 for SuddenR, 0.93 for InstaD and 0.95 for SlowD.