Strategy for stopping failure cascades in interdependent networks

TL;DR

This paper proposes a probabilistic reconnect strategy to prevent catastrophic failure cascades in interdependent networks, demonstrating increased resilience especially in low-degree networks through theoretical analysis and simulations.

Contribution

It introduces a novel reconnect-based mitigation strategy for failure cascades in interdependent networks, supported by percolation theory and simulation validation.

Findings

Resilience increases with higher reconnect probability γ.

Strategy is more effective in networks with low average degree.

Theoretical results align with simulation outcomes.

Abstract

Interdependencies are ubiquitous throughout the world. Every real-world system interacts with and is dependent on other systems, and this interdependency affects their performance. In particular, interdependencies among networks make them vulnerable to failure cascades, the effects of which are often catastrophic. Failure propagation fragments network components, disconnects them, and may cause complete systemic failure. We propose a strategy of avoiding or at least mitigating the complete destruction of a system of interdependent networks experiencing a failure cascade. Starting with a fraction $1-p$ of failing nodes in one network, we reconnect with a probability $\gamma$ every isolated component to a functional giant component (GC), the largest connected cluster. We find that as $\gamma$ increases the resilience of the system to cascading failure also increases. We also find that our strategy is more effective when it is applied in a network of low average degree. We solve the problem theoretically using percolation theory, and we find that the solution agrees with simulation results.