Leveraging Laplacian Pseudoinverse for Spreader Node Detection in Interdependent Human-Machine Networks

In this paper, we focus on identifying the best spreader node on a communication system composed by humandriven devices (e.g. smartphones) and machines like networked appliances, sensors, algorithms or bots. We model the system as an interdependent network consisting of a human network and a machine network coupled by interdependent links. Then, we define a novel metric of node importance in terms of information spreading, namely the Coupled Best Spreader (CBS), focusing on the concept of best spreader node in a network and pseudoinverse of the Laplacian of a graph. Through simulations on different network topologies, we show the effectivity of the CBS for predicting spreading patterns and maximizing/control diffusion in <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\mathbf{H 2 M}$</tex> systems. More specifically, analyzing the information dissemination dynamics from a a robustness perspective that involves the subsequent removal of the best spreader, we demonstrate that the attacks over the CBS nodes result in a rapid decrease of network robustness and the earlier fragmentation of the Erdős-Rényi and Watts-Strogatz based H2M networks, particularly causing phase transitions that disconnect human and machine components.