Research and innovation

Evidence-led R&D for secure autonomous infrastructure.

Our research programme focuses on problems that must be solved before AI agents can be trusted with consequential infrastructure workflows.

Research agenda

Four connected technical themes.

Agent Security

Adaptive Red-Blue Agent Systems

Constrained attacker and defender agents that evolve strategies through min-max learning, reinforcement signals and repeatable cyber-range evaluation.

Quantum-Safe Trust

Post-Quantum Agent Identity

Crypto-agile workload identity, signed actions, secure inter-agent communication and migration testing for long-lived infrastructure systems.

Cross-Domain CNI

Resilience Digital Twins

Simulation and hardware-in-the-loop environments linking energy, water and transport dependencies for attack propagation and recovery studies.

Governance

Continuous Agent Assurance

Machine-readable policy, behavioural evidence, human-oversight metrics and automated assurance packs for high-risk agent deployments.

Prototype roadmap

From controlled testbed to operational pilot.

Each prototype is designed to produce software, data and demonstrable controls not only a research report.

P1

Secure Agent Gateway

Identity, policy enforcement, tool mediation, human approval and tamper-evident logging.

P2

Adversarial Agent Lab

Red-team and blue-team agents tested in a bounded infrastructure simulation environment.

P3

Quantum-Safe Identity Testbed

Hybrid cryptographic profiles, certificate lifecycle tests and signed high-risk agent actions.

P4

CNI Decision-Support Pilot

One domain-specific agent deployed with progressive authority and measurable assurance.

Research principles

Useful innovation must survive operational reality.

01

Classical baseline first

Quantum-inspired or hybrid methods must demonstrate a measurable benefit over strong classical approaches.

02

Safety by constraint

Testing occurs in bounded environments with clear permissions, operator override and reproducible scenarios.

03

Evidence over claims

Performance, security and governance assertions are tied to datasets, tests and traceable evaluation criteria.

Collaborate

Research partners and CNI test environments.

We welcome collaboration with infrastructure operators, universities, cyber ranges, hardware suppliers and assurance specialists.