Echo-Guard: Acoustic-based Anomaly Detection System for Smart Manufacturing Environments

Abstract

The Industrial Internet of Things (IIoT) provides intelligence to industrial systems by linking sensors and devices with computer systems and software. However, it also increases the attack surface and exposes industrial systems to various types of IIoT threats. Smart manufacturing environments, built based on IIoT, are also automated and unattended and must respond to physical threats (e.g., vandalism, destruction, theft, etc.) and cybersecurity threats (e.g., DoS, DDOS, backdoor, etc.). In this paper, we propose Echo-Guard, an acoustic-based anomaly detection system to protect smart manufacturing environments. The Echo-Guard records acoustic signals coming from machines in the smart manufacturing environment and converts them into spectrogram images. The spectrogram images are further classified using CNN to detect anomalies in machine motion sounds. Our evaluation, conducted in a smart factory environment, shows that Echo-Guard is effective, achieving 99.44% accuracy, confirming the possibility that machine motion sounds can be utilized to detect anomalies.