Abstract
Radio frequency fingerprint identification (RFFI) distinguishes wirelessdevices by the small variations in their analog circuits, avoiding heavycryptographic authentication. While deep learning on spectrograms improvesaccuracy, models remain vulnerable to copying, tampering, and evasion. Wepresent a stronger RFFI system combining watermarking for ownership proof andanomaly detection for spotting suspicious inputs. Using a ResNet-34 on log-Melspectrograms, we embed three watermarks: a simple trigger, an adversariallytrained trigger robust to noise and filtering, and a hidden gradient/weightsignature. A convolutional Variational Autoencoders (VAE) with Kullback-Leibler(KL) warm-up and free-bits flags off-distribution queries. On the LoRa dataset,our system achieves 94.6% accuracy, 98% watermark success, and 0.94 AUROC,offering verifiable, tamper-resistant authentication.