Adversarial networks model communication through a graph of edges where information is sent from sender to receiver using a code. This presentation discusses the scenario in which a hostile adversary can corrupt along a restricted subset of edges within a “claw” network structure. In particular, we examine claw networks in which the sender sends a symbol along 2n edges, with n edges in each claw. We examine the families of claws with varying n, and give results on communication capacity as the influence of an adversary grows.
