Methodology
Detecting and Attributing Submarine-Cable Latency Anomalies with Two-Signal Cross-Validation
Evgeny Korolev - GeoCables · Technical note, June 2026 · v1.1
The key contribution here is not the detector itself, but a two-signal cross-validation layer that grades each alert against two independent physical signals: a change in the autonomous-system path (AS-path) at the time of the event, and whether other probes whose route physically traverses the same cable corridor observe the same degradation.
Both signals reflect fundamentally different physical processes and are largely independent of one another.
Their combined analysis separates genuine topology-level cable incidents from routing changes and from single-vantage noise. The numbers below are deliberately undramatic: the observation window contained no large-scale cable break - and the method correctly declines to manufacture one.
The path of one measurement
Every event on this site starts with a single traceroute. Click a station to see what happens there and where you can watch it live.
Our measurement points around the world keep sending traceroutes to addresses in many countries. A trace records every node on the path and how long it took to answer.
See it live: live map →Each node on the path gets a country, a city and a network. That turns a list of addresses into a line on the map.
See it live: network radar →From the countries the path crossed we work out which submarine cable carries it. A path from Singapore to Marseille cannot avoid the ocean floor.
See it live: cable map →Every pair of measurement point and cable has a usual delay: the average of the last twenty measurements. Each new measurement is compared with it in two ways at once, by ping and by trace.
See it live: cable monitoring →If the delay doubles in both measurements, we re-check, and only after two confirmations out of three do we open an alert. A single spike does not become an alert.
See it live: cable incidents →An alert is not yet a fault. We check whether other points see the slowdown and whether the route changed, and label it honestly: a widespread event, a route change or a false positive.
See it live: events feed →1. Data Foundation
| Asset | Scale |
|---|---|
| Submarine cables (with landing-point geometry) | 716 |
| Landing points (geo-coded) | 1,932 |
| Backbone / cable segments | 26,053 |
| Measurement network | own servers and an external network |
| Completed health checks (since 2026-03-01) | 168,699 |
| Cables under active measurement | 691 |
The method runs on a curated topology graph and a continuously accumulating measurement archive. Raw per-measurement RTT is preserved in append-only fashion (append-only - the immutability principle: new records are added, old ones are never deleted or modified - like a ship's log that guarantees the integrity of the historical record), so any future detector can be re-run over the full history.
This property cannot be reconstructed retroactively from public cable maps alone.
2. Detection (Baseline + Haversine Attribution)
Each check pings (and traceroutes) a target near a cable landing point.
A measurement is treated as a candidate anomaly when its RTT rises materially above that route's adaptive baseline (the baseline is the statistically “normal” latency computed from measurement history - like a body’s normal temperature: a deviation is flagged only when it is genuinely significant).
Candidates pass through a staged filtering funnel before any alert is raised - the first layer of false-positive suppression:
| Stage | Meaning | Count |
|---|---|---|
spike |
single measurement above the baseline | 624 |
anomaly_confirmed |
spike confirmed by recurrence or corroborating data | 189 |
alert |
promoted to a tracked incident | 114 |
Only ~18% of raw spikes (114 of 624) reach alert status.
Cable attribution is geometric: the latency jump is associated with the nearest cable segment by haversine distance (the haversine formula - a classical navigation method for computing the shortest distance between two points on the surface of a sphere; used by 19th-century navigators and today a staple of GPS and geospatial analysis) between the suspected hop and the candidate cables' landing points.
3. Two-Signal Cross-Validation
3.1 Signal A - AS-Path Reroute
When a submarine cable degrades, traffic frequently reroutes around it, changing the autonomous-system path.
An Autonomous System (AS) is an independently managed network or group of networks under a single administrative authority - such as a large ISP or cloud operator.
The entire internet comprises roughly 80,000 such systems, which coordinate routes via BGP - the “diplomatic language” spoken between networks.
For each alert we compare the AS-path before the event (the modal path for that probe→target pair) with the path at the event.
| Verdict | Meaning | Count |
|---|---|---|
route_change_break |
AS-path changed, with a large further latency rise on the new path | 4 |
route_change |
AS-path changed | 5 |
same_path |
RTT rose, path unchanged - consistent with congestion | 57 |
no routing history |
insufficient routing history | 48 |
During algorithm calibration, of the 66 alerts with sufficient routing history, 9 (13.6%) were independently corroborated by a measured AS-path change.
A naïve fingerprint over the raw IP-path is far too noisy: ECMP load-balancing (ECMP - Equal-Cost Multi-Path routing: when multiple equally valid paths exist to a destination, traffic is distributed across them randomly - like cars spreading across highway lanes) and intermittent timeouts produce ~18 distinct IP-paths per probe→target pair.
The signal only becomes stable on the AS-set fingerprint (~1.5 distinct per pair), which is what we use.
3.2 Signal B - Segment-Aware Multi-Probe Consensus
If only the detecting probe sees a degradation, the key question is whether the other probes are silent witnesses or simply not on the affected cable. A probe routing around the cable is not a witness - its silence proves nothing.
This is the classical logical trap: absentia probationis non est probatio absentiae - absence of evidence is not evidence of absence.
We therefore count a probe as an eligible witness only when its actual AS-path geo-traverses the same cable corridor as the alerting probe's path.
This geo-corridor check demoted 30% of the naïvely assumed same-cable “witnesses” as off-corridor.
| Verdict | Meaning | Count |
|---|---|---|
widespread |
majority of corridor witnesses also degraded - real cable event | 1 |
mixed |
some corridor witnesses degraded | 4 |
routing_event_non_cable |
alerting probe rerouted, corridor witnesses healthy → BGP/peering event, not a cable cut | 7 |
probe_specific_likely_fp |
corridor witnesses healthy, no reroute → local artifact | 44 |
narrow_path_event_possible |
single-probe cable - cannot rule out a narrow event | 1 |
insufficient_witness_context |
no concurrent corridor witness - status unknown | 57 |
3.3 Confidence Ladder
Both signals are largely independent: an AS-path change reflects a topological event in the network - a physical disruption in the routing path - whereas multi-probe consensus reflects the geographic breadth of degradation - how widely the problem has spread. These are fundamentally different physical processes that need not co-occur. Very few alerts satisfy both criteria simultaneously. Combining them:
| Tier | Definition | Count (of 114) |
|---|---|---|
| Dual-confirmed | AS-path change and multi-probe corroboration simultaneously | 0 |
| Single-signal | corroborated by exactly one of the two signals | 14 |
| Defensible false positive | corridor witnesses healthy and no reroute | 44 |
| Unclassified (coverage) | no witness present on the corridor - a probe-coverage limit, not method uncertainty | 56 |
The dual-confirmed count is zero, and that is the correct result: the March-June 2026 window contained no large-scale submarine-cable break.
A genuine major cut would light both signals simultaneously - just as an earthquake registers on multiple independent seismographs at once.
The value of the method lies in its ability to discriminate: confidently labelling 44 alerts (38%) as defensible false positives and isolating 7 events as routing changes that are not cable faults - rather than producing a dramatic body count on a quiet month.
4. How many alerts turned out to be false
All numbers refer to the period from March to June 2026.
- Out of 624 latency spikes, 114 became alerts. The other 510, that is 82%, were filtered out earlier: the spike did not repeat and nothing else confirmed it.
- Out of 114 alerts, 44 turned out to be false, that is 38%. Both checks came back negative for them: the route did not change, and other probes on the same corridor noticed nothing. These alerts are not sent in user notifications but stay visible in the dashboard.
- 14 alerts were confirmed by one of the two checks. None was confirmed by both at once: there were no major cable breaks in this period.
- For 56 alerts there is no answer. There was no second probe on the right corridor to confirm or refute the alert. This is a shortage of measurement points, not a weakness of the method: once a second probe appears, such alerts are judged by the same rule.
- Why an alert seen by a single probe is not automatically counted as false. A probe whose path does not use this cable is not a witness: its silence proves nothing. The simple rule "seen by one probe, so the alert is false" would have counted about 45% more false alerts than there really are.
5. Limitations
- Probe geography. A small fixed fleet of our own probes plus points of an external measurement network means many cables have no second vantage on the relevant corridor; for those, the consensus signal is structurally unavailable. Expanding the probe fleet directly removes the structural cause of most unclassified cases.
- Geo-resolution. Corridor matching is at country granularity from AS-path geolocation; some networks mis-geolocate, which we treat as unknown rather than mismatch.
- Attribution is geometric. Cable assignment is by haversine proximity; the corridor cross-check confirms it where path data exists (64 of 66 cases confirmed), but it is a heuristic, not a claim of ground truth.
- Quiet observation window. The reported period saw no major cable break. The method is validated by its discrimination, not by detecting a disaster - much like a smoke detector whose reliability is tested with test smoke rather than an actual fire.
6. Reproducibility & Data
Detection uses an adaptive threshold calibrated to each route's baseline distribution rather than a fixed multiplier; the consensus tiers reflect the share of eligible corridor witnesses that corroborate the degradation.
The exact parameterisation is intentionally omitted here and is available in a forthcoming complete publication or on request.
Raw per-measurement RTT and AS-paths are retained, so the classifier can be re-run end-to-end over the full archive under any revised method. Source signals: (ping + traceroute), own probes, and the curated cable/landing-point graph.
Historical series: country signal charts for events before 6 November 2026 are built on open series from the IODA project. Since 6 November 2026 events are determined by GeoCables' own measurements.
Method and figures may be cited with attribution. Live monitoring: Cable Health Monitor.
Figures are from the production system as of 2026-06-17 and will evolve as the archive deepens.
Vessels nearby
From 6 Aug 2026 to 8 Nov 2026 we compared 79 cable incidents on 57 cables with ordinary days on the same cables. A vessel that stopped within 5 km of the route (at anchor, moored or drifting slowly) was there in the 48 hours before 61% of the incidents. In ordinary 48-hour windows on the same cables such a vessel is present 57% of the time.
Ratio 1.13, 95% confidence interval 0.99 to 1.25: vessels near the route are not more frequent before incidents than on ordinary days, the difference is within the margin of error.
For the 28 significant incidents alone: 54% against 52%.
If you see our measurements in your logs
GeoCables is an independent monitor of submarine cables and internet outages. To see how networks are actually connected, our measurement hosts send traceroutes and pings to public addresses around the world. If you run a network and found these packets in your logs, this section is for you.
What we send
- Traceroute and ping (ICMP and UDP) to public router and server addresses.
- TLS handshakes and DNS lookups to a short list of well-known websites, to check whether they are reachable from different countries.
- Nothing else: no port scans, no login attempts, no payloads, no vulnerability probes.
How often
A typical address receives a traceroute from one of our hosts a few times a day, in most cases not more than once an hour. A small set of very popular destinations is measured every ten minutes. One traceroute is about twenty small packets.
How to opt out
Write to contact@geocables.com with the addresses or networks you want excluded. We add them to our exclusion list and stop measuring them, normally within one working day. You do not need to explain why.
What happens to the data
We publish aggregates: latency between cities, routing changes and outages. The open dataset does not contain IP addresses of intermediate routers.