pith. sign in

arxiv: 2607.01484 · v1 · pith:HKLOB43Inew · submitted 2026-07-01 · 💻 cs.NI · cs.AI

Fully Unsupervised Detection of Physical Contacts on Subsea Cables via State-of-Polarization Monitoring

Agastya Raj , Alvaro Doval , Tian Tian , Steinar Bj{\o}rnstad , Marco Ruffini This is my paper

Pith reviewed 2026-07-03 18:12 UTC · model grok-4.3

classification 💻 cs.NI cs.AI
keywords subsea cablesstate of polarizationunsupervised anomaly detectionphysical contact detectionfiber optic monitoringDSVDDtrawler damage
0
0 comments X

The pith

An unsupervised detector trained only on raw polarization data can rank all confirmed physical contacts on a subsea cable among its top anomalies.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper demonstrates a fully unsupervised approach to detecting physical contacts on subsea cables using state-of-polarization monitoring. By training a Fast-Slow DSVDD model on unlabeled recordings from a deployed cable, the method identifies all five known trawler contacts within the top 13 out of more than 122,000 samples. This matters because subsea cables face frequent damage from fishing and anchoring, and labeling every potential event is impractical for continuous monitoring. A sympathetic reader would see this as a way to automate anomaly detection in large-scale fiber optic networks without manual intervention.

Core claim

We present a fully unsupervised Fast-Slow DSVDD detector for continuous State-of-Polarization monitoring on a deployed subsea cable. Trained without event labels, it ranks all five confirmed trawler contacts within the top 13 of 122,174 recordings and surfaces additional corroborated cable-contact events.

What carries the argument

The Fast-Slow DSVDD, a two-component deep support vector data description model that learns to describe normal polarization states from unlabeled data and flags deviations as potential contact events.

If this is right

  • The detector requires no labeled examples of contacts or other events to operate effectively.
  • All known trawler contacts appear in the highest-ranked anomalies, suggesting high recall for this class of events.
  • Additional events flagged by the model have been corroborated as likely contacts, indicating it can discover new incidents.
  • Continuous monitoring becomes feasible on operational cables since the training uses only the normal data stream.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • Similar unsupervised methods might apply to detecting other rare events in optical fiber networks, such as earthquakes or equipment failures.
  • Operators could use the ranked list to prioritize physical inspections or repairs on long cable routes.
  • Extending the model to multiple cables or combining with location data could enable automated mapping of contact risks along the route.

Load-bearing premise

The polarization anomalies caused by physical contacts are distinct enough from normal variations and unrelated disturbances that an unsupervised model can separate them without any labeled examples.

What would settle it

Running the same detector on a new dataset from the same or similar cable where several confirmed contacts do not rank in the top 20 anomalies would falsify the claim.

Figures

Figures reproduced from arXiv: 2607.01484 by Agastya Raj, Alvaro Doval, Marco Ruffini, Steinar Bj{\o}rnstad, Tian Tian.

Figure 1
Figure 1. Figure 1: Experimental Setup labelled baselines [14], but both evaluate on con￾trolled testbed scenarios rather than continuous deployed data. Automated detection on deployed cables has been demonstrated in terrestrial set￾tings [7, 15, 16] but relies on supervised training. In this work, we transition from manual and su￾pervised approaches to fully unsupervised event detection on continuous long-duration SoP data f… view at source ↗
Figure 2
Figure 2. Figure 2: Detection framework of the Fast-Slow DSVDD model. The input SoP recording is processed by fast and slow heads, with four dilated convolutional branches. Branch-wise DSVDD distances are fused into a single recording-level anomaly rank [PITH_FULL_IMAGE:figures/full_fig_p002_2.png] view at source ↗
Figure 4
Figure 4. Figure 4: Logged and newly surfaced cable-contact events. Panel (a) shows a logged trawler contact, while panels (b)–(d) show three additional high-ranked candidates from the Fast-Slow DSVDD model. In each panel, the left subfigure shows the 30 s SoP waveform and the right shows the cropped DAS waterfall. Clear DAS signatures appear for the 3 June and 17 August events, while none is observed for the 10 June event wi… view at source ↗
read the original abstract

We present a fully unsupervised Fast-Slow DSVDD detector for continuous State-of-Polarization monitoring on a deployed subsea cable. Trained without event labels, it ranks all five confirmed trawler contacts within the top 13 of 122,174 recordings and surfaces additional corroborated cable-contact events.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 0 minor

Summary. The paper presents a fully unsupervised Fast-Slow DSVDD detector for continuous State-of-Polarization (SoP) monitoring on a deployed subsea cable. Trained without event labels on the unlabeled data stream, the detector ranks all five independently confirmed trawler contacts within the top 13 of 122,174 recordings and surfaces additional corroborated cable-contact events.

Significance. If the reported ranking holds, the work would provide empirical evidence that unsupervised anomaly detection on SoP streams can surface physical contact events at high rank in a real deployed system, offering a label-free approach to subsea cable monitoring with potential operational value. The use of a large real-world dataset and the concrete top-k ranking of known events are strengths that distinguish it from purely synthetic evaluations.

major comments (2)
  1. [Abstract] Abstract: the central empirical claim (all five confirmed contacts ranked in top 13 of 122174) is load-bearing for the paper's contribution, yet the abstract and available text supply no implementation details on the Fast-Slow DSVDD architecture, feature extraction from SoP time series, training procedure on the unlabeled stream, or anomaly-score threshold, preventing assessment of whether the ranking could arise from confounds such as diurnal variations or other polarization sources.
  2. [Abstract] Abstract: no data statistics (e.g., recording duration, sampling rate, number of independent cable segments) or verification procedure for the 'additional corroborated cable-contact events' are provided, making it impossible to judge whether the unsupervised detector's output is robust or sensitive to the specific choice of DSVDD hyperparameters.

Simulated Author's Rebuttal

2 responses · 0 unresolved

Thank you for the opportunity to respond to the referee's report. We address the major comments point by point below. We believe the full manuscript provides the necessary details, but we will revise the abstract to make them more prominent as suggested.

read point-by-point responses

  1. Referee: [Abstract] Abstract: the central empirical claim (all five confirmed contacts ranked in top 13 of 122174) is load-bearing for the paper's contribution, yet the abstract and available text supply no implementation details on the Fast-Slow DSVDD architecture, feature extraction from SoP time series, training procedure on the unlabeled stream, or anomaly-score threshold, preventing assessment of whether the ranking could arise from confounds such as diurnal variations or other polarization sources.

    Authors: The full manuscript details the Fast-Slow DSVDD architecture in Section 3, feature extraction in Section 2.2, the training procedure in Section 4.1, and the anomaly scoring in Section 4.2. We acknowledge that the abstract is concise and will expand it to include a high-level description of these components to allow readers to better evaluate potential confounds. This revision will be made in the next version. revision: yes

  2. Referee: [Abstract] Abstract: no data statistics (e.g., recording duration, sampling rate, number of independent cable segments) or verification procedure for the 'additional corroborated cable-contact events' are provided, making it impossible to judge whether the unsupervised detector's output is robust or sensitive to the specific choice of DSVDD hyperparameters.

    Authors: Data statistics including recording duration, sampling rate, and cable segments are provided in Section 2.1 of the manuscript. The verification procedure for additional events is described in Section 5.2, where we discuss corroboration with external sources. Regarding sensitivity to hyperparameters, we include an ablation study in the appendix. We will update the abstract to summarize these aspects for completeness. revision: yes

Circularity Check

0 steps flagged

No significant circularity; empirical ranking result is self-contained

full rationale

The paper presents an empirical result from applying a fully unsupervised Fast-Slow DSVDD detector to an unlabeled State-of-Polarization stream. The central claim is that this detector ranks five externally confirmed trawler contacts within the top 13 of 122174 recordings. No derivation chain, mathematical prediction, or fitted parameter is shown to reduce to its own inputs by construction. The method trains on the data stream itself without event labels, and performance is evaluated against independent confirmation of contacts; this is a standard unsupervised anomaly detection benchmark with no self-definitional, self-citation load-bearing, or ansatz-smuggling steps visible in the provided abstract and claim description. The result stands or falls on whether polarization anomalies are empirically distinct, which is an external falsifiable question rather than an internal definitional loop.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

No free parameters, axioms, or invented entities are described in the abstract.

pith-pipeline@v0.9.1-grok · 5580 in / 1109 out tokens · 29737 ms · 2026-07-03T18:12:03.883096+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

18 extracted references · 18 canonical work pages

  1. [1]

    Damage to submarine cables from dragged anchors

    International Cable Protection Committee. “Damage to submarine cables from dragged anchors”. ICPC Viewpoints, updated 24 February 2025, accessed 21 April 2026, International Cable Protection Com- mittee. [Online]. Available: https : / / www . iscpc . org / publications / icpc - viewpoints / damage - to - submarine-cables-from-dragged-anchors/

    work page 2025

  2. [2]

    First Im- pact Movement Characterization of Shallow Buried Live Subsea-Cable

    S. Bjørnstad, K. S. Y amase Skarvang, D. Roar Hjelme, A. Tunheim, F . Fjermestad, and E. Østerli, “First Im- pact Movement Characterization of Shallow Buried Live Subsea-Cable”, in2024 Optical Fiber Communications Conference and Exhibition (OFC), Mar. 2024, pp. 1–3

    work page 2024

  3. [3]

    Experience from Long-term Monitoring of Subsea Cables using Distributed Acoustic Sensing

    O. H. Waagaard, J. P . Morten, E. Rønnekleiv, and S. Bjørnstad, “Experience from Long-term Monitoring of Subsea Cables using Distributed Acoustic Sensing”, en, in27th International Conference on Optical Fiber Sensors, Alexandria, Virginia: Optica Publishing Group, 2022, Th2.4.DOI:10.1364/OFS.2022.Th2.4

    work page doi:10.1364/ofs.2022.th2.4 2022

  4. [4]

    Analysis of saturation effects of distributed acoustic sensing and detection on signal clipping for strong motions

    C.-R. Lin, S. von Specht, K.-F . Ma, M. Ohrnberger, and F . Cotton, “Analysis of saturation effects of distributed acoustic sensing and detection on signal clipping for strong motions”, en,Geophysical Journal International, vol. 241, no. 2, pp. 971–985, Mar. 2025.DOI: 10.1093/ gji/ggaf089

    work page 2025

  5. [5]

    Multi-Modal Fiber Sensing for Offshore Environmental and Infrastructure Monitoring

    K. Alexoudis, F . Azendorf, A. Doval, and S. Bjørnstad, “Multi-Modal Fiber Sensing for Offshore Environmental and Infrastructure Monitoring”, en, 2026

    work page 2026

  6. [6]

    Sensing earthquakes in aerial single- and multi-core fiber optic networks

    K. S. Vaskinn et al., “Sensing earthquakes in aerial single- and multi-core fiber optic networks”, en,Jour- nal of Optical Communications and Networking, vol. 18, no. 4, B119, Apr. 2026.DOI:10.1364/JOCN.584845

    work page doi:10.1364/jocn.584845 2026

  7. [7]

    A Smart Sensing Grid for Road Traf- fic Detection Using Terrestrial Optical Networks and Attention-Enhanced Bi-LSTM

    F . Usmani et al., “A Smart Sensing Grid for Road Traf- fic Detection Using Terrestrial Optical Networks and Attention-Enhanced Bi-LSTM”, en,Journal of Lightwave Technology, vol. 43, no. 10, pp. 4624–4634, May 2025. DOI:10.1109/JLT.2025.3543180

    work page doi:10.1109/jlt.2025.3543180 2025

  8. [8]

    Risky event classification leveraging transfer learning for very limited datasets in optical networks

    K. Abdelli, M. Lonardi, J. Gripp, S. Olsson, F . Boitier, and P . Layec, “Risky event classification leveraging transfer learning for very limited datasets in optical networks”, en,Journal of Optical Communications and Networking, vol. 16, no. 7, p. C51, Jul. 2024.DOI: 10.1364/JOCN. 517529

    work page doi:10.1364/jocn 2024

  9. [9]

    First observations of subsea cable physical contacts combining State of Polarization (SoP) and Distributed Acoustic Sensing (DAS)

    S. Bjørnstad, A. Doval, and A. Tysdal, “First observations of subsea cable physical contacts combining State of Polarization (SoP) and Distributed Acoustic Sensing (DAS)”, en, 2026

    work page 2026

  10. [10]

    Vision Transformers for Anomaly Clas- sification and Localization in Optical Networks Using SOP Spectrograms

    K. Abdelli et al., “Vision Transformers for Anomaly Clas- sification and Localization in Optical Networks Using SOP Spectrograms”, en,Journal of Lightwave Tech- nology, vol. 43, no. 4, pp. 1902–1914, Feb. 2025.DOI: 10.1109/JLT.2024.3519755

    work page doi:10.1109/jlt.2024.3519755 1902

  11. [11]

    Deep Learning for Detection of Harmful Events in Real-World, Noisy Optical Fiber Deployments

    L. Sadighi, S. Karlsson, C. Natalino, L. Wosinska, M. Ruffini, and M. Furdek, “Deep Learning for Detection of Harmful Events in Real-World, Noisy Optical Fiber Deployments”, en,Journal of Lightwave Technology, vol. 43, no. 13, pp. 6092–6101, Jul. 2025.DOI: 10.1109/ JLT.2025.3557748

    work page arXiv 2025

  12. [12]

    Low-complexity SOP-based vibration broadband sensing and efficient recognition for stable IM/DD optical interconnects in data centers

    B. Y ang et al., “Low-complexity SOP-based vibration broadband sensing and efficient recognition for stable IM/DD optical interconnects in data centers”, en,Jour- nal of Optical Communications and Networking, vol. 17, no. 8, p. 692, Aug. 2025.DOI:10.1364/JOCN.559810

    work page doi:10.1364/jocn.559810 2025

  13. [13]

    Ex- perimental Demonstration of Bending Eavesdropping Detection in Optical Communications Using a Physics- Informed Convolutional Network

    W. Qin, X. Gong, W. Hou, L. Gan, and L. Guo, “Ex- perimental Demonstration of Bending Eavesdropping Detection in Optical Communications Using a Physics- Informed Convolutional Network”, en,Journal of Light- wave Technology, vol. 44, no. 5, pp. 1636–1646, Mar. 2026.DOI:10.1109/JLT.2025.3647694

    work page doi:10.1109/jlt.2025.3647694 2026

  14. [14]

    ML- Based State of Polarization Analysis to Detect Emerging Threats to Optical Fiber Security

    L. Sadighi, S. Karlsson, C. Natalino, and M. Furdek, “ML- Based State of Polarization Analysis to Detect Emerging Threats to Optical Fiber Security”,IEEE Transactions on Network and Service Management, vol. 23, pp. 432– 442, 2026.DOI:10.1109/TNSM.2025.3607022

    work page doi:10.1109/tnsm.2025.3607022 2026

  15. [15]

    Overview on the state of polarization sensing: Application scenarios and anomaly detection algorithms

    S. Pellegrini et al., “Overview on the state of polarization sensing: Application scenarios and anomaly detection algorithms”, en,Journal of Optical Communications and Networking, vol. 17, no. 2, A196, Feb. 2025.DOI: 10. 1364/JOCN.537881

    work page 2025

  16. [16]

    Prompt Once, Manage All: A Unified LLM Framework for Multi-Task Optical Link Management

    K. Abdelli, “Prompt Once, Manage All: A Unified LLM Framework for Multi-Task Optical Link Management”, en, Journal of Lightwave Technology, pp. 1–12, 2026.DOI: 10.1109/JLT.2026.3660177

    work page doi:10.1109/jlt.2026.3660177 2026

  17. [17]

    Deep one-class classification

    L. Ruff et al., “Deep one-class classification”, inProceed- ings of the 35th International Conference on Machine Learning, J. Dy and A. Krause, Eds., ser. Proceedings of Machine Learning Research, vol. 80, PMLR, Oct. 2018, pp. 4393–4402

    work page 2018

  18. [18]

    Polarization sensing of network health and seismic activity over a live terres- trial fiber-optic cable

    C. J. Carver and X. Zhou, “Polarization sensing of network health and seismic activity over a live terres- trial fiber-optic cable”, en,Communications Engineering, vol. 3, no. 1, p. 91, Jul. 2024.DOI: 10.1038/s44172- 024-00237-w

    work page doi:10.1038/s44172- 2024