Environment-Agnostic RF Fingerprinting through Protocol-Compliant IQ Imbalance Estimation
Abstract
Zero Trust Architecture (ZTA), which requires continuous and context-aware authentication, has emerged as a response to the growing vulnerabilities of traditional security mechanisms in dynamic networks. Radio Frequency (RF) fingerprinting, which extracts hardware imperfections, is a promising physical-layer technique for device identification and authentication within ZTA. However, existing RF fingerprinting approaches often suffer from poor robustness to environmental variations or require protocol modifications or specialized instruments, significantly limiting their practical deployment. In this work, we present a robust and environment-agnostic RF fingerprinting system enabled by a novel, lightweight, and protocol-compliant in-phase/quadrature (IQ) imbalance estimation algorithm. Our approach is rooted in a key insight: the transmitted data, once decoded, can be directly used to estimate device-specific IQ imbalance without specially crafted signals or additional instrumentation. By leveraging protocol-supported carrier frequency offset (CFO) estimation, a closed-form least squares solution, and a dual-channel deep learning architecture that jointly processes raw IQ data and extracted hardware-specific parameters, we enable robust and environment-agnostic device identification while maintaining full compliance with standard communication protocols. Extensive real-world indoor and outdoor experiments show that our system consistently outperforms state-of-the-art methods and achieves near-perfect identification accuracy, even in challenging wireless environments.