Spoofing is deception — a false signal that moves your position subtly off the truth. Jamming is denial — broadband noise that simply buries the signal under interference. The distinction matters operationally: a spoofer wants the ship to trust a wrong answer; a jammer wants it to distrust any answer. Both are now daily operational facts in the Black Sea, the eastern Mediterranean, the Gulf and the Red Sea corridor.
The AI detection problem is correspondingly different. Spoofing detection looks for a <em>wrong but plausible</em> position; jamming detection looks for the <em>absence</em> of usable signal plus the signature of interference. This article covers what the research shows for the denial side.
Key findings
- GNSS jamming is now documented as routine in multiple theatres — a denial problem as much as a deception one. Established
- AI detection of jamming/interference is easier than of spoofing in one sense (the signal disappears) and harder in another (the interferer has incentive to look like a fault, not an attack). Maritime-GPT analysis
- Meaconing — record-and-replay of legitimate signals — is the grey zone between the two and the hardest to distinguish from honest signal loss. Established
- The resilient-PNT literature converges on multi-source redundancy (terrestrial nav, inertial, celestial, radar-fixing) as the real answer; AI detection is the tripwire, not the fix. Established
- Jamming detection aboard has operational value chiefly as an early-warning that the position fix is degraded — prompting manual fix-taking and heightened lookout. Research demonstration
Denial vs deception — two different alarms
A spoofed position looks like a real position; the detector's job is to find the inconsistency. A jammed receiver loses lock entirely or reports garbage; the detector's job is to tell interference from equipment fault, antenna shadowing, or honest ionospheric noise. The failure mode that matters is <em>silent denial</em> — a receiver that stops warning and starts interpolating on a dead signal. That's the case AI detection is actually for. Maritime-GPT analysis
What detection methods exist
- <strong>Signal-strength / C/N0 monitoring.</strong> A sudden noise floor rise is the clearest jam signature — cheap, effective, and a feature of modern receivers. Established
- <strong>Spectrum anomaly ML.</strong> Classifying the interference type (broadband, swept, meaconing) from the received spectrum — demonstrated on RF datasets. Research demonstration
- <strong>Position-consistency checking.</strong> Where residual signals exist, cross-checking GNSS against radar ranges, echo-sounder, inertial — the same physics-check the spoofing detector uses. Research demonstration
- <strong>Multi-receiver correlation.</strong> Two receivers in different locations shouldn't degrade identically unless the interference is real and wide-area. Maritime-GPT analysis
Where detection ends and resilience begins
Detecting jamming is the easy part. What the ship does next is the hard part: revert to dead reckoning, radar-fixing, celestial, or terrestrial aids — none of which are automated conveniences. The resilient-PNT literature is clear that AI detection's role is to trigger the fallback, not to be the fallback. A ship that knows its GNSS is jammed and has no practiced secondary-fix procedure is only marginally better off than one that didn't know. Established
What we know
- GNSS jamming is routine in several operational theatres and detectable via signal monitoring.
- Denial (jamming) and deception (spoofing) need different detectors and different responses.
- AI detection's real value is the early-warning that drives manual fix-taking.
What we don't
- How reliably AI distinguishes jamming from receiver fault or antenna shadow at sea.
- Whether any merchant fleet has standardised an AI-driven jamming-response drill.
- The true rate of meaconing vs simple jamming in the field — poorly characterised.
Being tested
- Spectrum-ML classifiers for interference type.
- Multi-source PNT fusion with automatic degradation alerts.
- Onboard interference mapping and reporting networks.
Maritime-GPT analysis
Jamming detection is the more honest problem because it admits what it is: an alarm, not a fix. The ship that loses GNSS doesn't need a smarter detector — it needs a drilled fallback to radar, DR and terrestrial aids. AI's value is the minutes of warning it buys, and the confidence that a silence is an attack rather than a fault. Buy the tripwire; invest in the fallback. Maritime-GPT analysis
This is a research feature, not operational guidance. Simulation results are reported as simulation; demonstrations are not deployments; company claims are attributed as such. Verify safety-critical decisions against official nautical publications and flag-state requirements.
Sources & evidence
IMO's framing of GNSS interference and resilience as a safety issue.
The deception side of the same problem — spoofing vs jamming.
The resilience requirements under which jamming response falls.