Cybersecurity Risk Assessment Using LLM Agents and Graph Data Science
summary
LLM agents ingest RSS STIX feeds, PDFs and websites, extract IoCs (malicious IPs, file hashes, log alarms), and use planning + reflection to validate them; filtered intel maps to a CMDB stored in Neo4j, then graph algorithms (degree centrality, PageRank, shortest path) compute node scores and predict risk propagation across three Palo Alto firewalls. A demo finds seven VAPT vulnerabilities, ranks RDP at 9.8 priority, and highlights two vulnerable workstations, producing patch order and visual graphs for SIEM/SOAR integration as an operational output.
AI-written summary · about
chapters (10)
- 0:00Introduction and agenda
- 1:47AI maturity in security
- 4:43Cognitive architecture patterns
- 7:07Modern SecOps workflows
- 10:10Graph databases for security
- 12:07Threat knowledge graph design
- 14:10Security co-pilot demonstration
- 18:39Graph data science algorithms
- 23:02Backend technical execution
- 26:57Summary and Q&A
related talks
31mSix Degrees of Infiltration
30mNavigating the New AI Security Landscape
48mHunting Stealth Adversaries with Graphs & AI
20mAttack of the Graph: Visual Tools for Cyber Analysis
27mAgentic Cyber Defense with External Threat Intelligence
26mCyRAG: Retrieval Augmented Generation for Robust Cyber Defense
22mEnabling Security Defenders with AI
44mTL;DR: Applying AI to Security