← Home
// Product family

Agentic security

Controlling and defending AI agents. EAS Defense, Elyne Red to challenge it, and Lab Director to run the laboratory.

Flagship product · Elyne Corp

EAS Defense

EAS controls actions routed through its interfaces: identity, capabilities, signed policy and evidence. Managed mode adds a restricted Windows profile for the launched process, to be qualified for the target software.

Deterministic decisionsSigned policyWindows confinementRed Unified included
Guided technical evaluationArchitecture and evidence →
Elyne Red UnifiedEvaluation

Exercise systems and AI agents within an authorised scope.

Included with EAS Defense, Red Unified orchestrates MITRE ATT&CK and ATLAS scenarios. Actions have real effects on the target; an action’s technical outcome and the defence verdict remain separate measures.

  • Campaign manifest: targets, transports, duration and capabilities.
  • Stop and restoration procedures defined before execution.
  • Red Agentic: the specialisation for AI-agent interfaces.
› Technical details
Windows dev13 is aligned with the EAS R45 kit. Linux dev14 was checked on Ubuntu WSL with SSH/SFTP and host-key verification; no EAS Linux defence campaign is qualified. Windows R21B concerns an earlier clone: 43 of 52 steps defended, with 9 inconclusive, and does not qualify R45.
Elyne Red AgenticTraining

The AI-agent specialisation of Red Unified.

Red Agentic targets AI-agent interfaces and flows in authorised exercises. It helps test protections and prepare teams for MITRE ATLAS scenarios.

  • Scenarios targeting agentic interfaces.
  • Actions and effects observed on the authorised target.
  • Technical outcomes separated from defence verdicts.
› Technical details
Red Agentic is part of Red Unified, not an additional standalone engine. Campaign manifests define permitted targets and capabilities. Target-agent confinement and defence effectiveness must be qualified using each campaign’s observations and evidence.
Elyne Lab DirectorLaboratory

Run the laboratory, from preparation to reporting.

ELD checks prerequisites and scope, follows approved steps and gathers evidence. Red executes offensive actions; ELD manages the laboratory. It remains distinct from EAS Defense and Red Unified.

  • Preflight checks, campaign journal and report.
  • Deterministic workflow, with optional local-model assistance.
  • Incomplete outcomes remain explicitly incomplete.
› Technical details
The operator creates and restores the encrypted VMware snapshot, followed by checks. Automatic closure after restoration is not established in demonstrations G/H; trial H remains RESTORING. A generated report or zero exit code alone does not establish a defence PASS.

Let’s discuss your project.

Contact us