Leveraging Graph AI and GPUs to Win the US Cyber Command AI Challenge

Discover how to leverage graph AI and GPUs to revolutionize cybersecurity investigations and detection engineering in this 53-minute conference talk from NDC Security in Oslo. Learn about the innovative methods and tools that led to a first-place finish in the US Cyber Command's Alert AI Challenge, including the use of graphs, vector embeddings, and GPU techniques. Explore how combining these technologies into a single pipeline resulted in a 97% reduction in alert volumes for ISPs and universities, while creating more effective situational awareness views for prioritization and interactive investigations. Gain insights into the application of LLMs and AI for automating graph workflows in zero-trust identity monitoring and investigations. Understand how to approach event analysis differently, utilizing modern graph AI techniques such as hypergraphs, UMAP, graph neural networks, GPU ETL, and GPU feature engineering. Get hands-on experience with free and open-source tools like Jupyter Notebooks, Nvidia RAPIDS, PyGraphistry[AI], DGL/PyG, and cu_cat to enhance your own cybersecurity workflows.

Leveraging graph AI and GPUs to win the US Cyber Command AI challenge - Leo Meyerovich