Identification of IoT Devices Based on Hardware and Software Fingerprint Features

Unauthenticated device access to a network presents substantial security risks. To address the challenges of access and identification for a vast number of devices with diverse functions in the era of the Internet of things (IoT), we propose an IoT device identification method based on hardware and software fingerprint features. This approach aims to achieve comprehensive "hardware-software-user" authentication. First, by extracting multimodal hardware fingerprint elements, we achieve identity authentication at the device hardware level. The time-domain and frequency-domain features of the device's transient signals are extracted and further learned by a feature learning network to generate device-related time-domain and frequency-domain feature representations. These feature representations are fused using a splicing operation, and the fused features are input into the classifier to identify the device's hardware attribute information. Next, based on the interaction traffic, behavioral information modeling and sequence information modeling are performed to extract the behavioral fingerprint elements of the device, achieving authentication at the software level. Experimental results demonstrate that the method proposed in this paper exhibits a high detection efficacy, achieving 99% accuracy in both software and hardware level identification.