Sweeping Up...
summary
AI-driven agentic garbage collector that reads free‑text names, descriptions and tickets to find contextual misconfigurations, claiming to spot temporary/testing/expired rule detection missed by heuristics. Built as a supervisor plus data‑retrieval, multiple detection agents and a judge, integrating GraphQL, Jira, Confluence, DocuSign and time/event tools and using Claude 3.5 via Amazon Bedrock as the LLM backbone. Presenter cites Gartner’s 50% stat, 3.5M unfilled roles, thousands of firewall rules and deployed product integrations, reports higher detection quality and much faster discovery than annual audits but offers no independent quantitative validation.
AI-written summary · about
chapters (12)
- 0:00Introduction
- 1:05Misconfigurations in critical systems
- 7:21Agentic AI engine for security
- 9:33Semantic analysis for config issues
- 10:38Temporary rules in firewall policies
- 13:55Testing rules and their risks
- 17:10Expired rules with contextual analysis
- 24:30Contextual misconfiguration crumbs
- 25:15Agentic framework building blocks
- 35:43Supervisor orchestration
- 39:02AI as proactive security sweeper
- 40:15Q&A
related talks
30mDetection Allegro: Composing Detection Rules with... (Raphael Ruban, Chen Cao)
27mCatch Me If You Can: Hunting Misconfigurations Before They...
32mCybersecurity meets Generative AI: Automating Your Compliance...
26mSecurity Automation For GraphQL
19mFuzzing in the GenAI Era
44mUnlocking the Future: AI is the Key to CISOs Top Challenges
32mAutomating Web Application Bug Hunting
39mBuilding Safe AI Agents