Smart Secure Authentication Sensing for Maritime Traffic: Sensor Individual Recognition Perspective

Security of shipborne automatic identification system (AIS) sensor equipment in maritime traffic is not only essential for navigational safety but also crucial for global economic and social development. Sensor individual recognition (SIR) is a popular method for maritime sensor authentication, which involves extracting radio frequency fingerprint (RFF) information that is difficult to tamper with to identify sensor equipment, providing an additional layer of security for maritime sensor networks. Existing deep learning (DL) SIR methods often rely on a large amount of high-quality labeled data to guide the model in learning RFF features, thus achieving high-performance models. However, under noncooperative communication conditions, it is difficult to obtain a large amount of high-quality labeled sensor signal data, and manually annotating sensor signals is extremely time consuming. Therefore, for the realistic scenario with limited labeled data, this article introduces semi-supervised learning (Semi-SL) theory, combines encoder with perturbation consistency regularization semi-supervised methods, uses unlabeled data to reduce the model’s dependence on labeled data, and employs virtual adversarial training (VAT) to approximate the real label distribution of unlabeled data, thereby improving the model’s recognition performance in scenarios with few labeled samples. Experimental results show that the proposed method achieves an identification accuracy of over 85% on a dataset containing signals from 50 classes of shipboard AIS sensors, even with only ten labeled samples per class. This performance surpasses that of pseudolabeling and the teacher-student structured Semi-SL methods.