Path 18: AI & AI Security — KashSec Academy

How LLMs work, prompt injection, OWASP LLM Top 10, AI red teaming. 23 free lessons, advanced level, starting with "How LLMs actually work"

Level: Advanced · 23 lessons · Free, no account needed.

Lessons in this path

  1. How LLMs actually work — A friendly primer without the mystique.
  2. Transformer Architecture, Deep Dive — Attention, context windows, and why prompt injection runs so deep.
  3. Why AI security is new — What makes AI attacks unlike any before.
  4. AI Threat Modeling — Systematically mapping an AI system's real attack surface.
  5. Prompt injection — The SQL injection of AI — direct and indirect.
  6. Indirect Prompt Injection, Deep Dive — Full attack chains and the dual-LLM defensive pattern.
  7. Jailbreaks & bypasses — Bypassing policy, not just instructions.
  8. Jailbreak Defenses & Guardrails — Classifiers, self-critique, and layered protection.
  9. OWASP LLM Top 10 — The shared checklist for AI application security.
  10. RAG Security Risks — Access control, poisoning, and the retrieval-trust gap.
  11. Poisoning & adversarial ML — Backdoors, extraction, adversarial examples.
  12. Multimodal Attack Vectors — Image, audio, and file-upload injection surfaces.
  13. AI red teaming — Methodology, harms taxonomy, ethics.
  14. AI Safety Evaluation & Benchmarking — Attack success rates, benchmarks, and rigorous measurement.
  15. Your AI-security career — The 12-month plan for a new frontier.
  16. Defensive Prompt Engineering — The cheapest real mitigation available to any builder.
  17. Agentic AI Risks — As AI agents become more powerful, new risks emerge.
  18. LLM Attack Taxonomy — A clear map of the different ways attackers try to break AI systems.
  19. LLM Governance & Compliance — Managing AI systems responsibly at scale.
  20. LLM Red Teaming Playbooks — Structured approaches to systematically test LLM applications.
  21. LLM Supply Chain Attacks — Attacks that target the components and data used to build and run LLMs.
  22. Practical AI Red Teaming Labs — Hands-on challenges you can do safely with free or low-cost LLMs.
  23. Securing LLM Applications — How to build and deploy LLMs more securely.