(POSTER) Beyond RFF - Context Integrity Monitoring via Image-Based RF Analysis

In constrained wireless Internet of Things (IoT) systems, physical-layer signals are influenced by both transmitter hardware impairments and variations in the surrounding propagation environment. Existing approaches typically consider these effects separately, either focusing on Radio Frequency Fingerprinting (RFF) for device authentication or using the Channel State Information (CSI) for device-free sensing. In this paper, we investigate the use of raw In-Phase - Quadrature (IQ) samples, typically used for RFF, to also perform context integrity monitoring, without requiring explicit channel estimation, pilot signaling, or protocol modifications. We formulate such a problem as an anomaly detection task, and we operationalize it through an autoencoder analyzing images built from In-Phase - Quadrature (IQ) samples at the receiver of the RF link. Experimental results obtained on a software-defined radio testbed using BPSK-modulated signals demonstrate reliable detection of transmitter deviations and human-induced context perturbations, up to 7 meters from the transmitter.