Vertex-connectivity for node failure identification in Boolean Network Tomography

We study the node failure identification problem in undirected graphs by means of Boolean Network Tomography. We argue that vertex-connectivity plays a central role. We prove bounds on the maximum number of simultaneous node failures that can be identified in arbitrary networks. We argue that (augmented) grids are a class of networks with large failure identifiability, and provide very tight results in this context.