Figure 3

Heterogeneous probability of uncovering links. We show the size of the giant component for heterogeneous likelihoods of nodes knowing their neighbors in the Dark Web network. We segregate nodes into those who are part of the network ‘core’ and that share illicit content versus the remainder of nodes that do not share content. We then consider different likelihoods of nodes having information on their neighbors based on whether the link is from a core node to a non-core node \(p_{\mathrm{core}\to \mathrm{reg}}\), from a non-core node to a core node \(p_{\mathrm{reg}\to \mathrm{core}}\), or between two core nodes \(p_{\mathrm{core}\to \mathrm{core}}\). We show the giant component, S as a function of the likelihood that core nodes can identify regular node neighbors and consider different likelihoods that a regular node can identify a core node. We find a rapid increase in the fraction of nodes that can be identified as both \(p_{\mathrm{reg}\to \mathrm{core}}\) and \(p_{\mathrm{core}\to \mathrm{reg}}\) reach around 10%. We chose a fixed value of the likelihood that core nodes can identify other core nodes of 10%. Results are averages of 1800 simulations on the Dark Web network