Vector space embeddings and data maps for cyber defense
Benoit Hamelin, John Healy · 2024
Cyber defense, and in particular threat detection, requires gaining insight from very large amounts of telemetry data, using unsupervised learning. This talk presents a method for embedding such data in vector spaces and exploring it through interactive visualization and labeling, using data maps. We demonstrate this method by looking at the baseline behaviours of hosts in the open ACME3 dataset of host-based telemetry.