A Radio Frequency Fingerprint-Based Zero Trust Authentication System for IoT Networks

The ultimate goal of next-gen networks is seamless communication across all types of devices, systems, people, and machines. Given the chance for network-wide effect of a single hack, this interconnectedness raises issues about privacy and security. Regardless of the precise data transmitted, radio frequencies fingerprint (RF fingerprint) analysis is a crucial technique used to physically identify the distinct radio transmitter. The RF fingerprint represents the variations in radio transmitter internal components and includes extensive nonlinear properties. The extraction and identification of radio frequency fingerprints has been widely used to improve the safety of radio frequency communications in both civilian and military settings. In response to this threat, the Zero Trust (ZT) model has emerged as an important tool for protecting data privacy and network security. With an emphasis on resource-constrained IoT devices, this study offers EPS-CNN, a new framework for wireless device identification that utilizes deep learning. It is intended to function as the device authentication layer inside the ZT architecture. The heart of EPS-CNN is a Convolutional Neural Network (CNN) that uses a specific representation of RF signals called the Double-Sided Envelope Power Spectrum (EPS) to build a device identity. This EPS captures the hardware characteristics of the device and ignores information that isn't related to it. When evaluated in same-domain (day, setting, and channel) settings, the suggested structure achieves testing accuracy of over 99%. When tested in crossday and cross-location scenarios, it achieves 93% accuracy. By providing a groundbreaking approach to enable ZT IoT device identification, our research proves that the suggested framework is the best way to improve the precision, resilience, and flexibility of deep learning-based methodologies.