HexaSec investigates practical security problems across AI-enabled and high-trust systems, turning research into repeatable test methods, defensive controls and inspectable technical evidence.
HexaSec research concentrates on security questions that can become testable methods, controlled systems or defensible engineering practice.
Failure modes across LLM, RAG and tool-enabled systems.
Repeatable testing, deterministic control logic and change-driven regression.
Traceability, retained records, provenance and independently inspectable technical evidence.
Defensive research into adversarial behaviour and cyber-system responses inside controlled environments.
ACTIVE / EXPLORATORY ARE HEXASEC PUBLIC RESEARCH-POSITIONING LABELS — NOT FORMAL TRL CLASSIFICATIONS
Some work remains exploratory. Other research becomes tooling, methods or product capability once it is mature enough to test in practice.
Repeatable security regression testing for LLM, RAG and tool-enabled assistants using deterministic checks, explicit policy and retained evidence.
VIEW PRODUCTResearch into how execution context, scenario outcomes, policy decisions, traceability and integrity material can remain inspectable after an assurance run.
VIEW EVIDENCEExploratory research into LLM reasoning, graph intelligence and coordinated agents for modelling adversarial behaviour and defensive response inside controlled environments.
SIMULATION-ONLY / NO LIVE TARGET OPERATIONHexaSec research starts with a defined security question and progresses through controlled modelling, prototyping, testing and evidence before mature work is transferred into engineering.
PUBLIC POSITIONING MODEL / NOT A FORMAL SCIENTIFIC LIFECYCLE, TRL FRAMEWORK OR CERTIFICATION PROCESS
Define the security problem.
Represent the system, behaviour or threat.
Build the smallest useful technical implementation.
Exercise the idea inside controlled conditions.
Retain observable results and technical artefacts.
Move sufficiently mature work into engineering, tooling or product capability.
RESEARCH OUTPUT / METHODS · PROTOTYPES · ENGINEERING TRANSFER
PROJECT AURORA
Project AURORA explores how language-model reasoning, graph intelligence and coordinated agents can be used to model adversarial behaviour and defensive response inside controlled environments.
EXPLORATORY RESEARCH / NOT A SHIPPING PRODUCT
Acquisition + curation
Knowledge + structured representation
AI inference + simulation
Orchestration + control
Visualisation + analyst interaction
HexaSec research is intended to push difficult security questions far enough to become useful engineering while keeping experimentation bounded, inspectable and defensible.
Experiments remain inside defined environments and operational boundaries.
Research should produce methods and results that can be repeated and inspected.
Claims should be supported by observable technical artefacts rather than model self-assessment alone.
Mature research can become engineering methods, prototypes or product capability.
NOT ALL RESEARCH BECOMES A COMMERCIAL PRODUCT