Review the retrieval pipeline as an attack surface

1. 01

   Concepts & system design

   60 min· read · map the mechanismtimer

   R10 — OWASP SSRF & File Upload Cheat Sheets

   R39 — OWASP Threat Modeling

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

2. 02

   Project implementation

   120 min· build · integrate · debugtimer

   Map the entire RAG ingestion and retrieval pipeline as an attack surface: file uploads, URL fetchers, PDF/HTML parsers, object storage, embedding workers, and vector search.

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

3. 03

   Security lab

   60 min· attack your own systemtimer

   Threat-model SSRF, path traversal (../../), zip/archive bombs, oversized files, and parser crashes across the ingestion pipeline.

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

4. 04

   Verification & evidence

   30 min· test · record evidencetimer

   Complete RAG data-flow diagram and STRIDE threat model.

   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

• R10OWASP Cheat Sheet Series↗ cheatsheetseries.owasp.org

  Authorization, SQL Injection Prevention, File Upload, SSRF, Secrets Management, and Logging Cheat Sheets.

• R39OWASP Threat Modeling↗ owasp.org

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

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

• docsOWASP Cheat Sheet Series · reference checklist
  File Upload Cheat Sheet

  Lists allowlisted extensions, generated storage names, size caps and out-of-webroot storage, the exact controls you map onto the ingestion stage's upload boundary.

  R10checked 2026-11-11
• articlePrompt Security · research blog
  Poisoning RAG Pipelines via Vector Embeddings

  Describes embedding-level poisoning that survives chunk-level inspection, extending the attack-surface map past ingestion into the embedding worker and vector store.

  R10R04checked 2026-11-11
• videoSecureTechIn · 10 min
  AI Security Lesson 15: RAG Poisoning Explained for Security Teams

  Walks the four-step chain from target question to crafted poison text to retrieval to answer distortion, which is the flow you annotate across ingestion, storage and retrieval.

  R04checked 2026-11-11
• videoTryHackMe · 31 min
  How Attackers Exploit RAG Systems | AI Security | TryHackMe

  Demonstrates why retrieval is the exploitable trust boundary, with hands-on attacks on the document-to-context path you are diagramming.

  R04R10checked 2026-11-11
• docsOWASP Cheat Sheet Series · reference checklist
  Threat Modeling - OWASP Cheat Sheet Series

  Gives the four-question loop plus system-modeling and threat-identification steps you apply to upload, parser, object storage, embedding worker and vector search in sequence.

  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.