# AEGIS — Red Team Framework - **Document ID:** ARCH-45 - **Phase:** E.3 — Adversarial/Defensive Teams - **Status:** Draft for review (post four-reviewer discipline) - **Version:** 0.1 - **Date:** 2026-07-10 - **Owner:** Chief Security Architect --- ## 1. Purpose Codify how AEGIS operates a red team — internal + external contributors who probe AEGIS's own defenses. Complements Threat Simulation Framework (ARCH-39) with human-driven, unscripted engagements. ## 2. Engagement Types - **Scoped assessment** — bounded surface, weeks-long, defined objectives. - **Continuous chaos-red** — narrow, always-on probes (per ARCH-50 CAV). - **Assumption-challenge** — a red engagement targeting one specific architectural assumption to prove/disprove (mandate #1). - **Regression red** — reproducing a prior finding to verify it stays fixed. ## 3. Engagement Cycle 1. Scope + rules of engagement + safety limits + Auditor observer. 2. Reconnaissance (both internal knowledge and external surface). 3. Exploit attempts, findings logged with evidence. 4. Report with severity, affected assurance claims, recommended mitigations. 5. Debrief with Blue Team + Architecture Council. 6. Findings → tickets → Debt or fixes. ## 4. Rules of Engagement (mandatory) - Written scope + safety limits + kill-switch. - Tenant data isolation for external engagements (test tenants). - No exfiltration of real tenant data. - Auditor observer on every engagement. - Findings held under embargo until fix or acceptance. ## 5. Security Control Spec (mandate #2) - **Purpose.** Adversarial discovery beyond scripted testing. - **Threats Addressed.** Novel attack chains; assumption drift; ARCH-03 adversary-profile coverage gaps. - **Residual Risk.** Time-bounded engagement misses long-persistence attackers. - **Verification Method.** Findings tickets + regression tests + red-engagement report + trend metrics. - **Operational Cost.** External engagement + internal time (weekly for continuous; quarterly for major). - **Failure Mode.** Under-resourced engagement misses findings. - **Recovery Method.** Cadence adjustments + external supplementation. - **Removal Criteria.** Only if AEGIS ceases; red-team is permanent. ## 6. Trust Score Contribution Red-team engagement outcomes feed the Trust Score (ARCH-52) under `Verification` and `Threat Activity` dimensions. ## 7. Assumptions (hypothesis format) - **H-1.** *Quarterly major engagements + weekly narrow probes are sufficient to catch high-impact drift.* - Evidence: prior engagement finding rate; industry benchmarks. - Validation: measure "finding-density" trend; compare with incident causes. - Confidence: Medium. - Expiration: 12 months. - Review Date: annually. ## 8. Independent Architecture Review - **F-1.** *Cost.* Balanced by proportional scoping; external only quarterly. - **F-2.** *Insider bias in internal red team.* Rotation + external supplementation. ## 9. Adversarial Architect Review - **A-1.** *Red team compromised.* External engagement contract + auditor + segregation. - **A-2.** *Findings suppressed.* Auditor visibility; regression tests are contract. ## 10. Operational Reliability Review - **O-1.** Cadence sustainable; scope tuned. - **O-2.** Findings integrated into Debt Register + IR runbooks. ## 11. Self-Critique - **S-1.** *Continuous chaos-red narrow scope may leave breadth uncovered.* Complementary to quarterly major engagements. - **S-2.** *External engagement quality varies.* Vetting + reference customers. ## 12. First-Target Analysis and Redesign **Target:** finding-suppression pathway (internal). Response: findings ticket immediately; Auditor sees raw findings; regression tests block until closure. ## 13. Future Risks / Known Limitations / Out-of-Scope / Retirement - **Future Risks.** AI-assisted attacks may outpace human red team. - **Known Limitations.** Time-bounded engagements can't emulate unlimited-patience attackers. - **Out-of-Scope.** Adversary emulation of nation-state operations at full sophistication. - **Retirement Conditions.** Never — red team is a permanent function. ## 14. Decisions ### D-45-1. Continuous narrow probes + quarterly major engagements + external supplementation - **Reason.** Balance depth + cost + coverage. ## 15. Change Log - **0.1 (2026-07-10)** — Initial draft.