Create a unified attack taxonomy across projects

1. 01

   Concepts & system design

   60 min· read · map the mechanismtimer

   R05 — MITRE ATLAS matrix

   R39 — OWASP Threat Modeling

   R41 — OWASP Agentic AI

   Read the assigned sections; map the mechanism, threat model and trade-offs.

2. 02

   Project implementation

   120 min· build · integrate · debugtimer

   Build a Unified Cross-Project Attack Taxonomy spanning P1, P2, and P3: identity/auth, data ingestion, prompt/context, RAG retrieval, MCP/tools, memory/checkpoints, model/supply-chain artifacts, inference serving, privacy, economic DoS, and safety/misuse.

   Build, integrate, debug and improve the increment in the active project.

3. 03

   Security lab

   60 min· attack your own systemtimer

   Map every deliverable (D1–D12) and capability (C1–C8, C10, C12) to the unified attack matrix and verify test coverage exists for every cell.

   Test the abuse case, failure path, authorization bypass or resource limit.

4. 04

   Verification & evidence

   30 min· test · record evidencetimer

   Unified Cross-Project Attack Matrix and coverage gap report.

   Add or run regression tests; record the 9-point evidence record.

5. 05

   Deliberate practice & Socratic review

   30 min· explain aloud · update gapstimer

   Answer the 8 Socratic questions without notes: What? Why? How? What if? Why not? How can I prove it? Where can it break? What have I learned?

   Re-explain the hardest concept aloud, inspect one edge case, update the gap log.

6. total5 h

• R05MITRE ATLAS↗ atlas.mitre.org

  ML attack tactics/techniques: prompt injection, data poisoning, model extraction, and exfiltration.

• R39OWASP Threat Modeling↗ owasp.org

  Four-question framework & STRIDE: assets, trust boundaries, data flows, threat identification, and mitigation tracking.

• R41OWASP Agentic AI Resources↗ genai.owasp.org

  Agentic threat taxonomy: tool misuse, excessive agency, memory poisoning, inter-agent trust, and human oversight.

Read only the named sections needed for today's task. Reading sits inside the 60-minute concepts block; it is not extra time.

• docsMITRE
  MITRE ATLAS™

  The living tactic/technique matrix for AI systems, giving you the canonical IDs to hang prompt injection, data poisoning, model extraction and exfiltration cases off in the unified taxonomy.

  R05checked 2026-11-11
• videoTech Explored By Abhi
  The AI Threat Landscape — MITRE ATLAS & OWASP Top 10 | AI Security Ep 2

  Teaches how to classify any AI attack against ATLAS and the OWASP LLM list, which is exactly the crosswalk step for one taxonomy spanning P1, P2 and P3.

  R05R41checked 2026-11-11
• videoNetsec Explained · 20 min
  STRIDE Threat Modeling for Beginners - In 20 Minutes

  Applies STRIDE to real components so assets, trust boundaries and data flows get named before you map agent-specific threats onto them.

  R39checked 2026-11-11
• articleOWASP
  Threat Modeling

  Sets out the four-question framework (what are we working on, what can go wrong, what will we do, did we do well) that structures the taxonomy and its mitigation tracking.

  R39moved 2026-11-11
• repoGitHub
  OWASP/threat-dragon

  A free data-flow-diagram editor that auto-suggests threats per element and stores mitigations, so the cross-project taxonomy becomes an inspectable artifact rather than a spreadsheet.

  R39checked 2026-11-11

A lab is complete only when it produces evidence. Installing PyRIT, Garak, Presidio or TransformerLens is not completion. Saved on this device only.