To address the critical need for secure IoT networks, this study presents a scalable and lightweight Curriculum Learning framework enhanced with Explainable AI (XAI) techniques, like LIME, to ensure transparency and adaptability. The proposed model employs a novel neural network architecture utilized at every stage of Curriculum Learning to efficiently capture and focus on both short- and long-term temporal dependencies, improve learning stability, and enhance accuracy while remaining lightweight and robust against noise in sequential IoT data. Robustness is achieved through staged learning, where the model iteratively refines itself by removing low-relevance features and optimizing performance. The workflow includes edge-optimized quantization and pruning to ensure portability that could easily be deployed in edge IoT devices. An ensemble model incorporating Random Forest, XGBoost, and the staged learning base further enhances generalization. The results demonstrate 98% accuracy on CIC-IoV-2024 and CIC-APT-IIoT-2024 datasets and 97% on EDGE-IIoT, establishing this framework as a robust, transparent, and high-performance solution for IoT network security.

Maritime domain awareness increasingly relies on Automatic Identification System (AIS) data. Yet, traditional monolithic backends struggle to scale with rising message volumes and offer limited resilience, data sovereignty, or rapid deployability. This study addresses the challenges by introducing a novel federated learning and microservices architecture for distributed vessel tracking. Each end node trains local models on a proprietary AIS Maritime environment via Dockerized microservices: Data Pre-Processor, Client Trainer, Aggregator, Global Model Updater, and XAI Service, enabling independent scaling, fault isolation, and federated governance without exposing raw feeds. A global model is aggregated using FedAvg and served with a sub-second latency of 10 ms of aggregation and 406 ms of inference. Experimental evaluation on four benchmark AIS snapshots yields strong predictive performance (MAE = 0.31, RMSE = 0.39, R2 = 0.78) and demonstrates transparent feature attribution via SHAP. These results validate the proposed architecture’s capability to deliver accurate, low-latency energy predictions while preserving data sovereignty and cross-node consistency. This study lays the foundation for robust, interoperable maritime analytics by bridging microservice agility with federated intelligence and XAI.