Turning Screen Recordings into Intelligent, Cause-Aware Knowlege Graphs
KGC 2026
•
27m
Penetration testing sessions yield rich screen recordings of expert pentesters using Kali tools, running scans, exploiting vulnerabilities, and pivoting intelligently. These videos encapsulate master-level expertise—intuitive decisions, creative chains, and risk assessments—yet this knowledge remains trapped, unusable for training, refinement, or automation.
This talk introduces Causal Visions, an automated pipeline extracting and structuring pentester decision-making from recordings into cause-aware knowledge graphs.
Using multimodal analysis: CV/OCR tracks terminals, Burp Suite, browsers, commands, outputs, and interactions; temporal modeling sequences multi-window actions; LLMs derive meaningful triples.
The key innovation: causal enrichment via temporal precedence, pentester priors (e.g., nmap scan → open ports → exploit attempt), and lightweight discovery—inferring edges like "Port 445 open → SMB exploit → privilege escalation."
A live demo processes a real session, yielding a queryable graph explaining pivots or abandoned paths.
Applications: Faster junior training via expert playbooks, methodology standardization, automated recon tools, reusable attack templates.
Challenges like noisy sessions, intent ambiguity, data sensitivity persist, but multimodal advances enable real-time capture.
Causal Visions reveals the "why" in pentesting mastery, turning fleeting sessions into enduring intelligence.
Up Next in KGC 2026
-
Ontologists Should Take Their Own Med...
https://drive.google.com/file/d/1tS9o4_p5DqMcEjc0XwTdlXHumojeEvWk/view?usp=share_link
As an ontologist, I must shamefully admit using data structures that are NOT graphs. I stored phone contacts, weight lifting information, task lists, etc., in things like spreadsheets and bulleted word document...
-
When AI Meets Clinical Expertise
https://drive.google.com/file/d/1QyL21ATgJB83oA038dsJtB1ZaTmUZlKQ/view?usp=share_link
LLMs can simulate clinical reasoning, but they can't capture the nuanced, individualized judgment that defines effective autism intervention. This talk presents a neural-symbolic approach: an ontology and knowle... -
Stop Betting, Start Building: AI Futu...
https://drive.google.com/file/d/1gpPJPutlfJqb2xUH-P2-WqtB1BUImwhc/view?usp=share_link
The AI gold rush is on. Every boardroom is placing bets on agentic AI—autonomous systems that perceive, reason, and act without human hand-holding. The market is bullish. The slide decks are stunning. And almost...
