Cascading failures in anisotropic interdependent networks of spatial modular structures

TL;DR

This paper investigates how anisotropy and modular structure in spatial interdependent networks influence their robustness against localized failures, revealing that anisotropic networks are generally more resilient and that attack shape significantly impacts failure propagation.

Contribution

It introduces a study of cascading failures in anisotropic, modular interdependent networks, highlighting the effects of spatial anisotropy and attack geometry on system robustness.

Findings

Anisotropic networks show increased robustness against localized attacks.

Anisotropic attacks are more effective than isotropic ones, even on isotropic networks.

Spatial structure and attack shape critically influence failure dynamics.

Abstract

The structure of real-world multilayer infrastructure systems usually exhibits anisotropy due to constraints of the embedding space. For example, geographical features like mountains, rivers and shores influence the architecture of critical infrastructure networks. Moreover, such spatial networks are often non-homogeneous but rather have a modular structure with dense connections within communities and sparse connections between neighboring communities. When the networks of the different layers are interdependent, local failures and attacks may propagate throughout the system. Here we study the robustness of spatial interdependent networks which are both anisotropic and heterogeneous. We also evaluate the effect of localized attacks having different geometrical shapes. We find that anisotropic networks are more robust against localized attacks and that anisotropic attacks, surprisingly, even on isotropic structures, are more effective than isotropic attacks.