Resilience of epidemics on networks

Dan Lu, Shunkun Yang, Jiaquan Zhang, Huijuan Wang, Daqing LiarXiv 2016 · arXiv

Abstract

This paper explores the "resilience" of epidemics—their ability to bounce back to a steady state after a temporary period of suppression. Researchers discovered that even if an intervention is strong enough to theoretically eliminate a disease, the epidemic can still recover if the control duration is too short. The critical time for permanent extinction scales predictably with the network's diameter, offering a new quantitative framework for public health and cybersecurity.

The Problem of Epidemic Resilience

In classical epidemiology, models often assume constant parameters like infection rates. However, real-world interventions—such as vaccination campaigns, quarantines, or anti-virus updates—introduce temporary "perturbations" that lower the infection rate. The central challenge is determining how long a control measure must last to ensure an epidemic does not "bounce back" once the measure is lifted. This study defines this recovery capability as "resilience," a trait that is beneficial in ecosystems but detrimental in disease management.

Methodology

The researchers utilized the Susceptible-Infected-Susceptible (SIS) model, where individuals cycle between susceptible and infected states without permanent immunity. They implemented a three-phase control framework:

  1. Steady State: The epidemic spreads at a natural rate ($\lambda_1$) until reaching a stable density of infected nodes ($\rho$).
  2. Control Phase: A perturbation lowers the infection rate to $\lambda_2$ for a specific duration ($cd$).
  3. Recovery Phase: The control is removed, and the rate returns to $\lambda_1$.
Temporal evolution of infected density $\rho$ under varying control durations.

The study combined stochastic simulations with mean-field reaction equations to predict the density of infected nodes over time across various topologies, including a real-world Facebook social network, Erdős-Rényi (ER) random networks, and Scale-Free (SF) networks.

Key Insights

The "Bounce Back" Phenomenon

A critical finding is that even if the control infection rate ($\lambda_2$) is below the epidemic threshold ($\lambda_c$)—meaning the disease would eventually die out if the control were permanent—the epidemic can still recover if the control is lifted too early. This occurs because, in finite networks, the infection density may drop significantly but fail to reach absolute zero before the perturbation ends.

Network Topology and Scaling Laws

Resilience is deeply tied to network structure. In Scale-Free networks with a power-law exponent $\gamma < 3$, the epidemic threshold is effectively zero, making these epidemics incredibly resilient. For other networks, the "critical control time" ($cd_{max}$) required for eradication follows a power-law scaling relationship with the network's diameter ($d$): $cd_{max} \sim d^\alpha$. As the network diameter increases, the time required for effective control grows predictably.

Key Results

Recovery probability $P$ as a function of infection rate $\lambda_2$ and control duration $cd$.
  • Threshold Dynamics: In ER networks, a sharp phase transition exists. If the control duration $cd$ is less than the critical threshold $cd_{max}$, the probability of recovery ($P$) remains near 1.0. Once $cd$ exceeds this threshold, $P$ drops sharply toward zero, as shown in Figure 5.
  • Predictive Accuracy: Mean-field equations accurately matched simulation results, confirming that recovery probability can be calculated based on whether the infected density stays above a critical floor ($1/N$) during the control phase.
  • System Size Effects: The maximum control time $cd_{max}$ increases with the system size ($N$). In an infinite system, the epidemic would theoretically always be resilient unless the control is permanent.

Limitations and Future Directions

While the SIS model provides a robust baseline, it lacks the permanent immunity characteristic of the SIR (Recovered/Immune) model, which would likely alter resilience dynamics. Furthermore, the study assumes homogeneous control across the entire network. Future research could investigate "targeted resilience," exploring how systems respond when control measures are focused specifically on high-degree "hubs" within the network.

Summary updated

Abstract

Epidemic propagation on complex networks has been widely investigated, mostly with invariant parameters. However, the process of epidemic propagation is not always constant. Epidemics can be affected by various perturbations, and may bounce back to its original state, which is considered resilient. Here, we study the resilience of epidemics on networks, by introducing a different infection rate ${λ_{2}}$ during SIS (susceptible-infected-susceptible) epidemic propagation to model perturbations (control state), whereas the infection rate is ${λ_{1}}$ in the rest of time. Through simulations and theoretical analysis, we find that even for ${λ_{2}<λ_{c}}$, epidemics eventually could bounce back if control duration is below a threshold. This critical control time for epidemic resilience, i.e., ${cd_{max}}$ can be predicted by the diameter (${d}$) of the underlying network, with the quantitative relation ${cd_{max}\sim d^α}$. Our findings can help to design a better mitigation strategy for epidemics.
BibTeX
@misc{lu2016resilience,
  author = {Dan Lu and Shunkun Yang and Jiaquan Zhang and Huijuan Wang and Daqing Li},
  title = {{Resilience of epidemics on networks}},
  year = {2016},
  eprint = {1610.06064},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/1610.06064},
}
Figure