HySense: Hybrid Event Occurrence Detection Method for IoT Devices

Abstract

Integrating IoT devices into physical environments through automation can result in device state inconsistency (DSI) anomalies. These anomalies can be caused by malicious attacks or device malfunctions, making it crucial to detect them for the security and reliability of IoT systems. Physical fingerprint-based verification methods have proven effective in detecting these anomalies by verifying whether the device state changes have physically occurred. However, these methods have limitations when there is no event notification. We propose a hybrid event occurrence detection method called HySense to address this issue. This method uses time-frequency (TF) analysis to locate events in the time domain and identifies real events from noise by spectrum analysis. To evaluate the performance of HySense, we used a real-world power signal dataset called PLAID, which contains 1170 events from 12 different appliances. Our experimental results demonstrate that HySense enhances detection performance with minimal computational overhead.