Attack Campaign Planning

Objective

Define how adversarial behavior will be generated, attributed, isolated, and reproduced after benign baselines exist.

Why this matters

The evaluation corpus needs realistic malicious behavior with trustworthy provenance rather than labels inferred from coarse time windows.

Design considerations

• Agentic attack execution.
• Deterministic targets and resettable entities.
• Trial-root provenance and transcript capture.

Key decisions

• Agentic attack campaign selected.
• Attack execution remains gated on validated telemetry and baseline collection.