Five of our projects get talked about separately and were designed to work together. This is an attempt to explain why, without the usual language about ecosystems.

The common thread is provenance. Each one answers a version of the same question: can you prove where this came from, who was allowed to see it, and what happened to it on the way?

CITADEL: proving the code is the code

CITADEL handles licensing and software integrity for environments where a casual download is not an option. Activation is tamper-evident, validation is policy-aware, and the cryptography is ECDSA-based, with key handling through Azure Key Vault. Offline, cloud-linked and hybrid activation all work, because real deployments are rarely all one or the other. The point is a verifiable link between what was written and what is running.

SIGIL: classification without relying on memory

SIGIL applies UK government classification metadata automatically across OFFICIAL, SECRET and TOP SECRET, in Microsoft Purview and SharePoint Online, aligned to JSP 604 and JSP 441. Most classification failures are not malicious; someone was busy and the marking did not get applied. Doing it structurally, so caveats propagate with the file, removes the opportunity for that mistake.

SHACE: encryption at the edge of the network

SHACE runs on hardened small-form hardware and does selective packet encryption, FIDO2 seed generation and post-quantum-ready key exchange. It bridges Azure Service Bus and AWS messaging through a clean and dirty pipeline, so traffic is sanitised on the way between a field node and command infrastructure rather than trusted on arrival.

BASTION: defending infrastructure that cannot be switched off

BASTION brings zero-trust network access, SIEM and SOAR orchestration and continuous monitoring together on top of Microsoft Sentinel and Elastic Security. Critical national infrastructure rarely gets a maintenance window, so detection, isolation and recovery have to work while the thing stays running. Red-team work keeps it honest; reporting keeps it funded.

VANGUARD: making training measurable

VANGUARD is our consultancy work around Bohemia Interactive Simulations‘ Virtual Battlespace: scenario design, performance tuning, and getting telemetry out into Power BI and Azure Data Explorer. A serial that is observed but not measured teaches whatever the instructor remembers. Replay with geospatial precision and performance data turns an opinion into a debrief.

How they fit

Verified software is distributed inside classified domains, moved across encrypted links, defended by monitoring that assumes compromise, and validated in simulation before anyone relies on it. That is the chain, and each link closes a gap the previous one opens.

They are also deliberately modular. Most organisations have an estate already and no appetite for replacing it, so each capability stands alone and drops into air-gapped networks, hybrid clouds or coalition federations without the others.

Outside defence

The problems are not unique to defence. Licensing integrity matters to any software vendor; metadata-driven compliance matters to anyone running Purview; edge encryption matters wherever industrial devices talk to a cloud; and monitoring built for critical national infrastructure transfers directly to utilities and transport. Simulation analytics apply wherever people are trained on something expensive to get wrong.

These are working systems rather than concepts, and they can be taken individually or as a programme. If you want to see one of them properly, ask for a technical briefing rather than a brochure.