Recognition model for vessel synchronized multi-MMSI codes based on ensemble learning

To address the issue of vessel synchronized multiple Maritime Mobile Service Identity (MMSI) codes spoofing in Automatic Identification Systems (AIS), where a single vessel broadcasts multiple MMSI codes simultaneously, this study proposes a recognition model for vessel synchronized multi-MMSI codes based on ensemble learning. The methodology first constructs synchronized multi-MMSI trajectory pairs through similar trajectory screening and manual analysis. Key features including time difference, course deviation, speed variation, geospatial distance, and minimum temporal-spatial distance are subsequently extracted to generate the feature dataset. The model employs machine learning ensemble algorithms, utilizing Support Vector Machine (SVM), K-Nearest Neighbors (KNN), Extreme Gradient Boosting (XGBoost), and Random Forest classifiers(RFC) as base learners, with a Multilayer Perceptron (MLP) serving as the meta-learner. Experimental validation using AIS data from Hainan, China. It demonstrates that the proposed model achieves superior accuracy (0.96) and stability compared to individual classifiers. Notably, this approach requires no external data sources such as radar, relying solely on AIS information for effective detection. The model demonstrates significant practical potential for maritime surveillance applications.