2025 · pipeline

AIS dark-activity detector

A pipeline that flags vessels going dark in contested waters, with a reproducible scoring method.

Built with
  • Python
  • AIS
  • GeoPandas

Problem

Vessel tracks are noisy, and “going dark” — switching off AIS — is easy to spot once and hard to score consistently across a whole corridor and a whole year. Without a reproducible method, every analyst’s dark-activity map is a different map, and none of them can be checked.

Approach

The detector establishes a per-corridor baseline rhythm, then flags absences that depart from it, weighting each by location and duration. Every flag carries a score and the inputs that produced it, so the result is auditable rather than asserted.

  • Ingest and clean raw AIS, de-duplicating by vessel and time.
  • Build a baseline transit model per chokepoint.
  • Score absences against the baseline; persist scores with provenance.

Outcome

A scoring method others can run on their own AIS extract and get the same numbers — the point is not a single map but a shared way of making one.