How to Map Networked Devices Using Graph-Based AI to Preempt Cyber Threats
summary
Graph neural networks convert passive NetFlow, sFlow and event logs into a network graph, train a graph autoencoder on ~70 node features to produce low‑dimensional embeddings, cluster those embeddings with K‑means and HDBSCAN, and deploy on NVIDIA Morpheus for GPU‑accelerated inference. On ~10,000 assets this workflow finds that 91% intra‑cluster traffic lets analysts focus on 9% cross‑cluster flows, surfaces misconfigured NTP and unexpected high‑packet communications, and runs refreshes on a ~two‑week cadence.
AI-written summary · about
chapters (13)
- 0:00Introduction
- 2:34Asset management challenges
- 5:31AI in cybersecurity
- 6:44Graph neural networks
- 11:26Graph autoencoders
- 14:00Data input features
- 16:07Clustering methodologies
- 17:33Case study and insights
- 22:48Nvidia Morpheus deployment
- 24:12Pipeline architecture
- 25:33Demo and application
- 27:31Getting started
- 28:39Q&A session
related talks
45mConnecting the dots to go from tabular security incident data to behavioral graph understanding
53mLeveraging graph AI and GPUs to win the US Cyber Command AI challenge
20mAttack of the Graph: Visual Tools for Cyber Analysis
49mA Real-World Example of Using Network Behaviour to Detect Cyber-Compromise
56mGraph Analytic Systems
27mEnhancing Financial Analytics With Event-Driven Architectures and Graph Neural Networks
26mAdvancing Financial Fraud Detection with Graph Dbs
31mStreamlining Threat Hunting in Cloud Environments with Jupyter