Vessel Detection and Localization Using Distributed Acoustic Sensing in Submarine Optical Fiber Cables

TL;DR

This work addresses the need to protect submarine cables from accidents and sabotage by leveraging Distributed Acoustic Sensing along existing fiber for real-time vessel monitoring. It proposes a general DAS+ML pipeline that includes data-driven spectral features and two learning tasks—vessel detection and distance estimation—evaluated over a 10‑day field deployment on a 28 km North Sea cable. The approach achieves strong detection performance (F1 over 0.90) and robust distance estimates (mean absolute error around 141 m) and outperforms a beamforming baseline, demonstrating practical viability for cable-protection applications. By releasing the fully annotated dataset and providing a reproducible processing stack, the study sets a foundation for broader deployment, cross-site generalization, and future extensions such as speed, size, and type estimation, as well as multi-target tracking in maritime surveillance contexts.

Abstract

Submarine cables play a critical role in global internet connectivity, energy transmission, and communication but remain vulnerable to accidental damage and sabotage. Recent incidents in the Baltic Sea highlighted the need for enhanced monitoring to protect this vital infrastructure. Traditional vessel detection methods, such as synthetic aperture radar, video surveillance, and multispectral satellite imagery, face limitations in real-time processing, adverse weather conditions, and coverage range. This paper explores Distributed Acoustic Sensing (DAS) as an alternative by repurposing submarine telecommunication cables as large-scale acoustic sensor arrays. DAS offers continuous real-time monitoring, operates independently of cooperative systems like the "Automatic Identification System" (AIS), being largely unaffected by lighting or weather conditions. However, existing research on DAS for vessel tracking is limited in scale and lacks validation under real-world conditions. To address these gaps, a general and systematic methodology is presented for vessel detection and distance estimation using DAS. Advanced machine learning models are applied to improve detection and localization accuracy in dynamic maritime environments. The approach is evaluated over a continuous ten-day period, covering diverse ship and operational conditions, representing one of the largest-scale DAS-based vessel monitoring studies to date, and for which we release the full evaluation dataset. Results demonstrate DAS as a practical tool for maritime surveillance, with an overall F1-score of over 90% in vessel detection, and a mean average error of 141 m for vessel distance estimation, bridging the gap between experimental research and real-world deployment.

Figures (20)

• Figure 1: General location and bathymetry (cable location has been displaced for security considerations).
• Figure 2: AIS data update rate histogram for the used dataset.
• Figure 3: AIS vessel speed histogram for the used dataset.
• Figure 4: Comparison of bathymetry sources in the cable under study. Sea floor depth provided by the cable owner (orange trace) and EMODnet (green dotted line). The Fiber position (buried cable) is also shown (gray trace). Diff. EMODnet vs. owner measures depth differences between EMODnet and owner data (dark blue trace). Selected fiber-optic cable regions are shown in blue background (250 sensors range) and pink background (50 sensors range).
• Figure 5: Class instance distribution for different distance thresholds.