ARCHIVES

Maritime security faces significant challenges due to intentional Automatic Identification System (AIS) shutdowns, often associated with illegal activities such as smuggling, piracy, and unauthorized fishing. Existing AIS- based detection methods rely on predefined rules and supervised learning, limiting their ability to adapt to complex real-world scenarios and leading to high false positive rates. To address these challenges, this study introduces a Hybrid Self-Supervised Deep Learning (HSSDL) framework that integrates multi- modal sensor fusion, anomaly detection, and graph- based trajectory analysis for enhanced maritime surveillance. The proposed system leverages self- supervised learning to pre-train deep learning models on vast amounts of unlabeled maritime data, improving the model’s ability to detect anomalous vessel behavior. Additionally, multi-modal sensor fusion combines satellite imagery, radar data, and environmental information to track vessels independently of AIS signals, reducing reliance on a single data source. Graph Neural Networks (GNNs) analyze vessel movement patterns, identifying suspicious trajectories that indicate deliberate AIS shutdowns. Furthermore, an unsupervised anomaly detection module employs deep reinforcement. Learning and clustering techniques to differentiate between intentional shutdowns and legitimate system failures. To enhance decision-making, the system incorporates Explainable AI (XAI), providing transparent and interpretable alerts to maritime authorities. Designed for real-time implementation, this innovation significantly improves detection accuracy, reduces false positives, and enhances adaptability in maritime security operations. The proposed system offers a robust solution for coast guard operations, naval defense, and global maritime surveillance, making it a strong candidate for patent protection and real-world deployment in safeguarding international waters.