Securing Agentic AI With PyTorch: Threat Modeling & LLM Red Teaming in Practice
summary
Practical guide that applies MAESTRO-based AI Threat Modeling to PyTorch agent architectures, mapping specific risks (model hallucination rates, vector DB exfiltration, service-account privilege chains) to mitigations and test cases. Demonstrates use of OWASP LLM Top 10 and the OWASP LLM Testing Guide, including a live prompt injection CTF that compromises a vendor-approval agent by extracting the system prompt and chaining to a privileged agent. Recommends LM firewalls, AI bill-of-materials, capability-based risk scoring and red‑teaming workflows.
AI-written summary · about
chapters (8)
- 0:00AI security challenges
- 4:21EU AI Act compliance
- 9:04Threat modeling agentic AI
- 14:43MESTRO framework
- 20:10LLM attack vectors
- 24:24Red teaming AI attacks
- 27:17AI vulnerability risks
- 30:34Demo and CTF overview
related talks
34mLLM Security 101 - An Introduction to AI Red Teaming
25mThreat Modeling for AI Apps with Attacks as Code
42mSecuring AI Agents...
54mA Practical Guide to AI Threats & Mitigation Strategies
19mKeynote: Towards Building Safe and Secure Agentic AI
27mKeynote: Towards Building Safe and Secure Agentic AI
43mSecurity of GenAI & LLMs: A New Era of Ethical Hacking
18mCommit Secure AI: Embedding Security in Code, Pipelines, and Production