Merging Human Cyber Security Expertise with Generative AI
summary
Claims a working Retrieval-Augmented Generation pipeline that pairs a BERT-GRC classifier for policy-gap detection with a self-hosted RAG stack, using Chroma vector DB plus Hugging Face embeddings and LangChain and a local LLM (GPT-for-all) on EC2 G4dn hardware. Benchmarks: sub-10s responses and roughly <$500/month compute for a small team, 5 analysts ≈20 queries/day (~100k tokens/month). Delivers MITER-embed open source mapping of detections to MITRE, live Slack integration, and industry validation with multiple partners.
AI-written summary · about
chapters (14)
- 0:00Introduction
- 1:52AI in cyber security today
- 4:01AI need in security ops
- 6:09CISO concerns on AI
- 8:19RAG and security
- 12:33Building RAG step by step
- 17:07RAG security considerations
- 20:42Psychology of AI adoption
- 22:51MITRE techniques to detections
- 25:59Threat intel and CVE analysis
- 28:28Vulnerability prioritization
- 31:39GRC automation
- 37:18RAG future directions
- 38:20Summary
related talks
01h 07mEmpowering Security Teams with Generative AI
59mRAG Against the Machine
50mRetrieval Augmented Generation Beyond Toy Demos
21mHow RAG is Redefining AI Capabilities
28mImproving responses in generative AI applications
25mAdvanced RAG Systems: Scaling Retrieval-Augmented Generation in Production
26mBreaking RAG Systems: Exploiting Vulnerabilities & Hardening Your GenAI Applicatio...
20mRetrieval Augmented Generation for AI Apps