Empowering Security Teams with Generative AI
summary
Overview of concrete work: demos show tokenization and embedding with Transformers, a local stack using Ama + LightLLM and a Seer 7B model, and cloud calls to GPT‑4/GPT‑3.5; a RAG pipeline embeds MITRE ATT&CK data into a vector DB and returns relevant documents; an LLM agent uses function calling to pick Microsoft Graph endpoints, fetch evidence with an access token, and summarize incidents; experiments validate zero/few‑shot chain‑of‑thought and handle larger contexts up to 32,000 token sizes.
AI-written summary · about
chapters (18)
- 0:00Generative AI for security
- 2:09AI ML deep learning basics
- 9:31Model weights parameters embeddings
- 13:43Recurrent models text limits
- 20:31Transformers self-attention
- 24:00Foundation models LLMs
- 31:29Prompting system messages context
- 37:28Open-source LLMs Hugging Face
- 40:17Zero-shot few-shot learning
- 43:31LLM hallucinations knowledge limits
- 45:37Retrieval augmented generation
- 50:14Threat intel query agent RAG
- 53:25LLM agents reasoning decisions
- 56:35Zero-shot chain of thought
- 58:41React agents loops
- 1:00:06LLMs tools APIs incident response
- 1:02:31Microsoft Graph API agent
- 1:05:05API automation LLM reasoning
related talks
41mMerging Human Cyber Security Expertise with Generative AI
15mHarnessing the Power of Generative AI for Intelligent Applications
49mBen Hall & Barbara Fusinska
37mAGENTS OF S.E.A.L.E.D: AI Agentic Cybersecurity Framework
44mUnlocking the Future: AI is the Key to CISOs Top Challenges
43mSecurity of GenAI & LLMs: A New Era of Ethical Hacking
01h 10mPractical Applications of Generative AI: How to Sprinkle a Little AI in Your App
49mUsing Generative AI to Strengthen & Accelerate Learning