A Wi-Fi-based Lightweight Personnel Identification System for IoT Applications

In recent years, Wi-Fi-based human activity recognition (HAR) has emerged as a research hotspot. However, the issue of security assurance in Wi-Fi-based HAR systems remains underexplored. A Wi-Fi-based personnel identification system can share hardware and data acquisition modules with Wi-Fi-based HAR systems, providing security assurance for these systems. To be effectively deployed on resource-constrained router devices and to ensure continuous security without significantly impacting the Wi-Fi HAR experience, a Wi-Fi personnel identification system must achieve low resource consumption, low inference latency, and high accuracy. In this paper, we investigate a Wi-Fi personnel identification system based on knowledge distillation. Our system ensures high accuracy in personnel identification while significantly reducing the identification time, thereby providing security assurance for Wi-Fi-based HAR. Experimental results demonstrate that the proposed personnel identification system can identify individuals with a low latency of 132[Formula: see text]ms on performance-constrained router devices, achieving an average accuracy of 99.64%.