Abstract
The book delivers a rich set of foundations, state-of-the-art knowledge, new approaches and methods for the purpose of anomalies detection, maritime traffic analysis as well as risk and reliability assessment. It addresses relevant research problems at the intersection of maritime transport in global economies, reliability and risk assessment, and information systems and data processing. The book provides a theoretical overview of available maritime data sources and approaches for maritime data analysis, as well as a set of novel tools and methods for maritime data retrieval, fusion, and analysis. The proposed methods are evaluated on real-life AIS data, covering the entire world and more than 200 thousands of vessels, illustrating how they may be used for anomaly detection and risk assessment.
Introduction
Maritime transport handles roughly 80% of global trade by volume, making it the backbone of the global economy. However, this domain faces significant threats from piracy, illegal fishing, and environmental hazards. While modern vessels generate vast streams of Automatic Identification System (AIS) data, this information is often incomplete, inaccurate, or intentionally manipulated.
Existing maritime surveillance systems often suffer from "information overload" and a lack of integration with external data sources like ownership records or weather patterns. This research presents a scalable, data-driven framework for Maritime Domain Awareness (MDA) that utilizes big data architectures to process billions of AIS messages and transform them into actionable intelligence.
Methodology and Big Data Architecture
To handle the computational load of processing terabytes of data annually, the authors implemented their algorithms using Apache Spark and a Lambda architecture. This allows for both batch processing of historical data and stream processing of real-time updates, providing an order-of-magnitude speedup over traditional relational databases.
Key Analytical Methods
- MRRAM (Short-term Maritime Reliability and Risk Assessment Method): This method uses Bayesian Networks (BN) to model the probability of ship delays. It categorizes risk into three classifiers: ship-related (static features), voyage-related (dynamic environment), and history-related (past behavior).
- SPP (Ship’s Punctuality Prediction): A multi-component algorithm that predicts a vessel's route, creates a travel time profile, and adjusts for real-time congestion, geopolitical hazards, and weather extremes.
- Anomaly Detection: The authors developed a typology for anomalies, ranging from static issues (black-listed flags) to dynamic behaviors like loitering (unusually low speed on the high seas) and angle anomalies (sharp course changes).
Data Fusion and Heterogeneous Sources
The research relies on a massive fusion of diverse datasets:
- AIS Data: A global dataset from Orbcomm satellites containing over 1.39 billion messages from 425,166 unique vessels.
- Deep Web Sources: Data scraped and fused from MarineTraffic, the ITU MARS database, and various Memoranda of Understanding (MoU) for ship detentions and inspections.
- Environmental Data: Historical and forecast weather data from the Copernicus Marine Environment Monitoring Service (CMEMS).
Key Insights and Results
A critical finding of the study is the inherent unreliability of self-reporting systems; approximately 23% of vessels provided incomplete or "not available" ship type data. To address this, the authors emphasize context-specific normalcy, using location-specific speed maps to distinguish between normal operations and loitering anomalies.
Performance Metrics
- Punctuality Accuracy: The SPP method provided a more accurate Estimated Time of Arrival (ETA) than the ship's captain in 88.42% of cases.
- Risk Prediction: The MRRAM method correctly classified ship reliability with an accuracy between 73% and 84%.
- Computational Efficiency: The Spark-based approach was nearly 10 times faster than traditional MSSQL for anomaly detection tasks involving 1,000 vessels.
- Static Anomalies: Analysis revealed that nearly 20% of all tankers were registered under "Flags of Convenience," a major factor in assessing environmental and security risks.
Conclusion and Future Work
The proposed framework demonstrates that big data technologies can significantly enhance maritime security and logistics. Future work will focus on integrating more complex sea-state variables like tides and currents, improving trajectory reconstruction through interpolation algorithms, and developing real-time dynamic re-routing capabilities to suggest optimal paths in response to hazards.
Summary updated