MeshGraph is a graph convolutional node classifier for detecting compromised assets
inside a simulated enterprise network. Nodes carry host telemetry, while edges encode
observed communication. Compromise begins at sparse seeds and propagates stochastically
through the network.
The GCN is compared with a logistic-regression baseline that sees identical host
features but cannot use graph structure.
This is a transductive benchmark: the graph and all node features… See the full description on the dataset page:
https://huggingface.co/datasets/ARotting/meshgraph-enterprise-network.
