Real-time Interference Identification via Supervised Learning: Embedding Coexistence Awareness in IoT Devices

TL;DR

This paper presents a lightweight, real-time interference identification method for IoT devices using supervised learning, capable of distinguishing multiple wireless standards with high accuracy and low latency on COTS hardware.

Contribution

It introduces a novel on-device interference detection approach leveraging spectral features and manifold supervised classifiers suitable for resource-constrained IoT hardware.

Findings

Achieves 90%-97% burst identification accuracy in real environments.

Operates with sub-millisecond identification time on COTS hardware.

Effectively distinguishes multiple wireless standards even under heavy interference.

Abstract

Energy sampling-based interference detection and identification (IDI) methods collide with the limitations of commercial off-the-shelf (COTS) IoT hardware. Moreover, long sensing times, complexity and inability to track concurrent interference strongly inhibit their applicability in most IoT deployments. Motivated by the increasing need for on-device IDI for wireless coexistence, we develop a lightweight and efficient method targeting interference identification already at the level of single interference bursts. Our method exploits real-time extraction of envelope and model-aided spectral features, specifically designed considering the physical properties of signals captured with COTS hardware. We adopt manifold supervised-learning (SL) classifiers ensuring suitable performance and complexity trade-off for IoT platforms with different computational capabilities. The proposed IDI method is capable of real-time identification of IEEE 802.11b/g/n, 802.15.4, 802.15.1 and Bluetooth Low Energy wireless standards, enabling isolation and extraction of standard-specific traffic statistics even in the case of heavy concurrent interference. We perform an experimental study in real environments with heterogeneous interference scenarios, showing 90%-97% burst identification accuracy. Meanwhile, the lightweight SL methods, running online on wireless sensor networks-COTS hardware, ensure sub-ms identification time and limited performance gap from machine-learning approaches.