GAN-based anomaly detection algorithm for maritime communication data

The rapid digitalization of maritime communication systems has introduced new challenges for reliable and efficient anomaly detection across heterogeneous shipborne networks. Traditional centralized and rule-based monitoring approaches are unable to cope with the dynamic, distributed, and bandwidth-constrained nature of maritime environments. To address these issues, this paper presents a GAN-Based Anomaly Detection Algorithm for Maritime Communication Data (MC-GAN), which integrates generative adversarial learning with adaptive temporal encoding and communication-efficient synchronization. The proposed framework learns the generative manifold of normal maritime traffic and reconstructs deviations as anomalies in real time. Through the integration of adversarial reconstruction, temporal attention, and distributed synchronization, MC-GAN achieves both high detection precision and reduced communication cost. Experimental evaluations on synthetic AIS data and the real-world MarComNet-Real dataset demonstrate that MC-GAN attains an AUC of 0.976 and an F1-score of 0.949, outperforming six competitive baselines while reducing bandwidth usage by over 40%. These results highlight the framework’s potential for intelligent maritime surveillance and its scalability to distributed edge environments in the era of smart shipping and autonomous communication systems.