Roadmap

Detailed roadmap under embargo

The detailed roadmap (channel tier specifications, capacity tables, safe-depth analyses, defender posture, and per-format detection plans) is withheld from the public site pending peer review and coordinated disclosure.

Once the associated paper is publicly released, the full roadmap will be restored here.

Public summary

The project's roadmap covers four general areas:

1. Channel research. Studying how covert information can be carried inside neural network weights across formats and architectures.
2. Detection. Building scanners and statistical tests that defenders can use to surface weight-level covert channels.
3. Defender tooling. Designing cleansing and canonicalisation pipelines that strip covert content while preserving model utility.
4. Supply-chain analysis. Understanding how the distributed model ecosystem changes the threat model.

Specific milestones, capacities, depth ranges, and detector thresholds are part of the embargoed academic submission.

For the high-level project framing, see the Home page.