What comes next, in order.
Prove it on real incidents.
The first job is a backtest. We replay documented episodes of regional jamming and vessels changing identity through GHAST and check what it does with them.
Three questions decide it. Does it flag the gap or the identity change, with the evidence shown? Does it tell area jamming apart from one vessel being spoofed? How many wrong alerts does ordinary traffic produce, measured on real traffic?
The results get published with the method, whichever way they go, along with an open benchmark of labelled incidents that anyone can run their own detector against.
Close the loop with analysts.
Today the console runs on recorded data. Next it reads from the running pipeline, so an incident on screen is one the system flagged an hour ago.
Every verdict an analyst records feeds the precision figure on the Trust page, which appears once enough incidents have been reviewed. Model versions get a place there too, so a number always says which model produced it.
Check a track against physical evidence.
A reported position is a claim. Radar imagery from public satellites can show whether a ship was where it said it was, and whether it vanished or met another vessel.
Pair the two and "this track is implausible" becomes "this track is implausible and nothing was at that spot." Free imagery revisits most coasts every few days, so this starts as an after-the-fact check. It will not be live tracking.
Evidence someone else can rely on.
A report that only makes sense inside the console is of limited use. Each incident will export as one evidence pack: the source records with their hashes, the version of the pipeline that raised the flag, and the full log of what the agent did.
An insurer, an auditor or an investigator should be able to reproduce the finding from the pack, and a qualified expert should be able to put their name to it.
Data with clear terms.
The pipeline reads a live public AIS feed today, which suits development. Moving to licensed sources with clear terms for commercial use is part of getting ready for customers.
Where feed coverage is thin, we plan to add receivers of our own along the coast.
Beyond ships.
The check underneath GHAST is general: is this reported location and track physically believable? Ships are where it starts. The same check applies wherever a position is reported and then trusted.
- Ports and terminals. A small receiver that listens for jamming and spoofing and keeps an audit trail.
- Trucking and delivery fleets. Fake-location apps, jammers and route fraud.
- Aircraft position broadcasts. Mapping GPS interference for airlines and drone operators.
- Offshore infrastructure. Vessels loitering near cables, pipelines and platforms.
These are explorations. None is committed, and each has to earn its place against what the backtest shows.
No dates.
The backtest may change the order, so there are no dates here. There are no accuracy figures either, because none have been measured yet.
Open the console or read how it works.