JOINT RADIO FREQUENCY FINGERPRINTING AND DISTANCE CLASSIFICATION VIA MULTI-TASK LEARNING

This thesis presents the development of a Multi-Task Learning (MTL) model for Radio Frequency Fingerprinting (RFF) and transmitter-receiver distance classification. The proposed ResNet18-based architecture processes images generated from I/Q signals acquired using Software Defined Radio (SDR) devices. Experimental results demonstrate that a single network can jointly perform both tasks, achieving average accuracies of 99.51% for transmitter identification and 94.01% for distance classification, with slight improvements over the corresponding single-task models.