Thirty-One Zones, No Transmitter: What AIS-Based Spoofing Measurement Can and Cannot Establish
A new study of more than 367,000 vessels puts 31 persistent spoofing zones on the map and finds Red Sea anomalies five months before a grounding. The limits the authors place on their own method are the part practitioners should carry away.

What the study reports
A measurement study from Georgia Institute of Technology, reported from 29 September, puts a number on what the maritime sector has described anecdotally. Working from Automatic Identification System broadcasts covering “more than 367,000 vessels” collected from late November 2024 to early February 2025, the researchers looked for groups of independent ships showing physically implausible movement at the same time and place: speeds above 60 knots, positions on land, and displacements of more than five kilometres from the expected track [1].
The counts are specific: 17,936 anomalous episodes involving 2,663 vessels, in 31 persistent zones accounting for more than 31,000 hours of implausible navigation [1][2]. The zones span the Black Sea, eastern Mediterranean, Red Sea, East Asian waters, the Baltic, the Strait of Hormuz, the Canary Islands, the Gulf of Mexico and the Panama Canal [1].
One caveat: this desk could not retrieve the preprint, because the hosting service refused the request. The figures come from two independent accounts that agree and reconcile internally — 13 corroborated zones plus nine strongly evidenced gives the 22 the second account reports [1][2]. That is reassuring but not the same as having read the method; anyone relying on these numbers should obtain the paper.
AIS is a witness, not an instrument
The limitations matter more than the count because of what AIS is. The IMO requires it on “all ships of 300 gross tonnage and upwards engaged on international voyages, cargo ships of 500 gross tonnage and upwards not engaged on international voyages and all passenger ships irrespective of size”, and the system transmits “the ship’s identity, type, position, course, speed, navigational status and other safety-related information” automatically to shore stations, other ships and aircraft [7].
The position in that broadcast is not measured by AIS. It is what the ship’s own receiver derived from satellite navigation, repeated outward. When a fleet reports itself on dry land, AIS has faithfully relayed a lie told to the receivers — which makes it a good wide-area detector of interference and a poor instrument for characterising it.
The authors say exactly this. Their detection is “designed to find regional interference affecting multiple vessels”, missing smaller displacements, incidents under thirty minutes and single-vessel spoofing, and they describe their measurements as “a lower bound on maritime spoofing activity” [1]. The second account records the same constraint in different words: AIS analysis “cannot identify the transmitter, operator, or spoofing technology responsible and likely misses short-lived, localized, and single-vessel incidents” [2].
The classification tiers are the finding
Splitting 31 zones into 13 corroborated, nine strongly evidenced and nine inconclusive is the most useful thing in the work, and the part a brief drops first: a third of the clusters a wide-area method surfaces cannot be confirmed from the same data. That is not a weakness but the shape of the evidence this layer yields, and anyone building a risk picture from commercial AIS feeds inherits the same ceiling. Treating all 31 as confirmed overstates the record by nine.
Port Sudan, and what was claimed at the time
The study’s most quotable finding is temporal: anomalies near Port Sudan in January 2025, affecting at least twelve vessels, resemble the pattern associated with the MSC Antonia grounding five months later [1][2].
How that grounding was characterised at the time matters. The 304-metre, Liberian-flagged vessel went aground near Eliza Shoals, about twenty nautical miles north-west of Jeddah, on 10 May 2025, having sailed from Sudan [8][10]. Pole Star Global’s Captain Steve Bomgardner attributed it to interference, which “happens when a threat actor introduces fake signals which the GPS unit picks up as genuine, therefore giving the vessel’s crew an inaccurate picture of where they are” [8]. Context cited then included UK Maritime Trade Operations reports of interference affecting multiple Red Sea vessels on 9 May, and Windward data showing average position jumps rising from about 600 km in late 2024 to about 6,300 km in early 2025, with more than 180 vessels affected [8][10].
Other reporting was more careful, describing “strong speculation” about spoofing or jamming and noting no statement from the operator [10]. An analyst’s assessment and a causal finding differ, and the new study does not convert one into the other: it shows the conditions were present months earlier, which is a prior, not a cause.
Aviation gets closer to the emitter
The contrast with aviation shows the ceiling is the data source, not the analysis. Work by Sherman Lo and colleagues at Stanford and the University of Colorado Boulder, using crowdsourced Automatic Dependent Surveillance Broadcast data with multilateration, ground monitors and AIS, resolves spoofing into geometric signatures: circles, semi-circles and hendecagons two to eleven kilometres across, with apparent speeds that let them infer the target [3].
From that they draw conclusions AIS cannot support. Near Smolensk, apparent speeds of “approximately 200 kilometers per hour” suggest the spoofing is aimed at large uncrewed aircraft rather than airliners; circles of 60 to 400 metres in Myanmar point at small commercial drones; roughly 300-metre circles at about 110 km/h on the India-Pakistan border point at small military drones [3]. They also observe that “spoofing in all three region[s] had aircraft being spoofed throughout the day suggesting that they may be continuously on” [3].
Shipborne radio measurement closes the gap. A GNSS monitoring vendor, which sells detection products and so has a commercial interest, published Baltic measurements from a research vessel between June and October 2025 resolving four synchronised emitters — a GPS L1 spoofing generator plus lower-band, upper-band and broadband jammers — with availability falling to 83.5% [9]. That is evidence identifying technology, and it needs a receiver in the water.
The machinery already exists
None of this is a gap in international law. ITU Constitution Article 45 requires stations to be operated “in such a manner as not to cause harmful interference to the radio services or communications of other Member States”, and Article 47 obliges states to prevent “the transmission or circulation of false or deceptive distress, urgency, safety or identification signals” [6]. Radio Regulation 15.1 forbids “the transmission of false or misleading signals” and 15.37 requires an administration receiving notice to “promptly investigate” [6].
The enforcement record shows the limit. ICAO condemned interference from Russia and North Korea in October 2025 as a breach of the 1944 Chicago Convention, and the ITU Radio Regulations Board urged Russia in November 2025 to “immediately cease any source of harmful interference” [4]. The ITU, ICAO and IMO secretaries-general have jointly expressed “grave concern”, IMO’s Arsenio Dominguez warning that interference “could cause collisions and grounding” [5]. What the regime lacks is not rules but attribution, which AIS cannot supply.
What to take from this
- When a spoofing-zone count reaches you, ask which tier each zone sits in; a third here are unconfirmed from the study’s own data [1][2].
- Do not treat an AIS-derived finding as evidence about an emitter; it shows receivers were deceived, not by whom or what [2][7].
- If you need characterisation rather than detection, that means a receiver on the route [9].
- Record the provenance of causal claims about incidents: an analyst’s assessment is useful and is not a finding [8][10].
- Keep conventional navigation competence available, as the three UN agencies ask, and file interference reports through the existing mechanism [5][6].
- Expect the Red Sea, Hormuz, Black Sea and Baltic to stay persistent rather than episodic, and plan passage accordingly [1][4].
The study is a real contribution: it replaces anecdote with a bounded count and names the places. Its more durable gift is the restraint. A method that tells you a lie was told, over 31,000 hours and 31 zones, while being explicit that it cannot say who told it, serves a practitioner better than one that quietly infers the rest.
Sources
Every R3KONX article cites its primary material. 10 sources, in order of first citation. Links open the original publication.
- 367,000-Ship Study Finds Global GPS Spoofing, Red Sea Activity Before Grounding Hackread · 2026-10-04
- Study Identifies Coordinated GPS Spoofing Across Global Shipping Routes Mallory · 2026-10-04
- Global Incidents of Aviation Spoofing in 2024-2025 Detected with Automatic Dependent Surveillance Broadcast Lo, Liu, Ibrahim, Chen, Akos and Walter, Stanford University and University of Colorado Boulder · 2026-01-01
- GNSS Interference Now a Constant of Modern Conflict, SWF Annual Report Finds Inside GNSS, reporting the Secure World Foundation Global Counterspace Capabilities 2026 report · 2026-04-08
- Joint IMO-ICAO-ITU statement on protecting satellite navigation from interference International Maritime Organization · 2026-10-06
- Harmful Interference to the Radio Navigation Satellite Service International Telecommunication Union · 2026-10-06
- AIS transponders International Maritime Organization · 2026-10-06
- Pole Star Confirms GPS Interference Caused MSC ANTONIA Grounding gCaptain · 2025-05-15
- New GNSS Interference Report Released: Shipborne Measurements Near the Kaliningrad Border GPSPATRON (vendor of GNSS interference detection products) · 2025-11-01
- MSC Container Ship Stranded In Red Sea After Suspected GPS Spoofing Marine Insight · 2025-05-13
Researched and written by the R3KONX analysis desk from the cited primary material. The preprint itself could not be retrieved from this desk: the hosting service refused the request. Its figures are therefore reported at second hand, from two independent accounts whose numbers agree and reconcile arithmetically, and are attributed to those accounts rather than to the paper. Every other claim here is traced to a document that was read in full. Corrections to event@r3konx.asia.
