Source-linked AI summary
Traffic-driven Epidemic Spreading in Finite-size Scale-Free Networks
S. Meloni, A. Arenas, Y. Moreno
TL;DR
The paper addresses epidemic spreading in finite scale-free networks when contacts are generated by traffic rather than assumed to occur continuously across all links. It develops and analyzes a traffic-driven SIS framework under unbounded and bounded delivery, finding that traffic lowers epidemic thresholds while congestion limits incidence.
Problem
Conventional epidemic models treat transmission as reaction from infected nodes to all neighbors, overlooking temporally limited interactions shaped by network traffic.
Method
The paper combines analytical treatment and numerical simulations of a traffic-driven SIS process on scale-free networks with unbounded or bounded packet delivery.
Results
Traffic determines epidemic thresholds: higher flow lowers them, while bounded delivery produces congestion that bounds epidemic incidence.
Takeaways & Limitations
Epidemic spreading in scale-free networks depends on contact flows and the traffic-defined propagation pathways, not only on static network structure.
Abstract
from arXiv · showhide
The study of complex networks sheds light on the relation between the structure and function of complex systems. One remarkable result is the absence of an epidemic threshold in infinite-size scale-free networks, which implies that any infection will perpetually propagate regardless of the spreading rate. The vast majority of current theoretical approaches assumes that infections are transmitted as a reaction process from nodes to all neighbors. Here we adopt a different perspective and show that the epidemic incidence is shaped by traffic flow conditions. Specifically, we consider the scenario in which epidemic pathways are defined and driven by flows. Through extensive numerical simulations and theoretical predictions, it is shown that the value of the epidemic threshold in scale-free networks depends directly on flow conditions, in particular on the first and second moments of the betweenness distribution given a routing protocol. We consider the scenarios in which the delivery capability of the nodes is bounded or unbounded. In both cases, the threshold values depend on the traffic and decrease as flow increases. Bounded delivery provokes the emergence of congestion, slowing down the spreading of the disease and setting a limit for the epidemic incidence. Our results provide a general conceptual framework to understand spreading processes on complex networks.
I. INTRODUCTION
The paper argues that conventional reaction-based epidemic models overlook temporally limited interactions and introduces a traffic-driven SIS framework. It analyzes unbounded and bounded packet delivery, showing that traffic shapes epidemic thresholds and, under congestion, limits incidence.
- Motivation: Reaction-based models assume infected nodes transmit to all neighbors at each time step, overlooking that real nodes interact with only subsets of neighbors concurrently.The paper motivates traffic-driven spreading by noting temporal interaction constraints in social, biological, and technological networks.
- Approach: The study introduces a theoretical SIS approach in which contagion occurs through contacts created by transport rather than reaction events.Transmission is possible when connected partners use their links and eventually contact one another.
- Traffic scenarios: Two traffic scenarios are analyzed: unbounded delivery and bounded delivery of packets per unit time.These scenarios are intended to encompass most real traffic conditions considered by the paper.
- Main findings: With unbounded delivery, increasing traffic lowers the epidemic threshold; with bounded delivery, congestion bounds both the threshold and infection prevalence.Bounded delivery causes queues to accumulate when traffic exceeds maximal delivery capacity.
II. RESULTS AND DISCUSSION
The simulations construct uncorrelated and clustered scale-free networks and drive packet traffic through shortest-path or greedy routing. Epidemic transmission occurs only along links carrying packets, making traffic determine effective contact opportunities.
- Network construction: The study generates uncorrelated configuration-model scale-free networks and clustered small-world scale-free networks embedded in a hidden metric space.The clustered model uses a power-law expected-degree distribution and distance-dependent connection probabilities.
- Traffic model: Traffic creates p = λN packets per time step with random origins and destinations, routed by shortest-path delivery or a greedy algorithm.The simulations use both network topologies and routing strategies to model packet transport.
- Traffic-driven contagion: Epidemic transmission occurs only when two nodes exchange at least one packet, rather than continuously across every potential link.The infection is transmitted with probability β when an infected node sends a packet to a susceptible neighbor.
A. Unbounded Delivery Rate
In the unbounded-delivery setting, traffic-driven SIS spreading is modeled through packet-mediated contacts, making epidemic thresholds depend on traffic flow and algorithmic betweenness. The threshold falls as traffic increases, with theory and simulations separating absorbing and endemic regimes.
- Model and analytical framework: Packet-mediated contacts replace always-active neighbor transmission, so the effective spreading network changes with local traffic activity.Only links carrying packets can transmit infection, and contact rates depend on packets traversing nodes.
- Model and analytical framework: Traffic-driven SIS dynamics use algorithmic betweenness to represent flow pathways and heterogeneous mean-field equations to model infected-node densities.Infection probability depends on packet flow through nodes rather than connectivity alone.
- Model and analytical framework: The analytical treatment assumes uncorrelated networks and approximates the flow between degree classes, with numerical simulations confirming the theory despite these approximations.The uncorrelated-network assumption is explicitly acknowledged as an approximation because such networks do not exist exactly.
- Threshold behavior: The threshold separates an absorbing phase, where infection disappears, from an active phase, where infection remains endemic.The threshold is obtained from the condition for a non-zero stationary solution.
- Threshold behavior: The epidemic threshold depends on the first two moments of the algorithmic-betweenness distribution and decreases with traffic, eventually vanishing at very large traffic flow in finite systems.This contrasts with a finite-size threshold expected from the classical reactive-diffusive framework.
- Threshold behavior: For random routing, the effective critical value is (βλ)c = <k>^2/(<k^2>w), recovering the corresponding earlier result when w = <k>.The random-protocol calculation treats packets as noninteracting random walkers.
B. Bounded Delivery Rate
With finite delivery capacity, traffic can congest queues and bound epidemic incidence. The congestion regime depends on delivery-capacity heterogeneity relative to routing betweenness.
- Bounded Delivery Rate: Above the critical traffic rate λc, queues accumulate packets and eventually grow without limit as congestion spreads through the network.The traffic threshold is governed by the node with maximum algorithmic betweenness.
- Bounded Delivery Rate: For node capacities scaling as 1 + k_i^η, η controls how delivery capability is distributed relative to traffic concentration.Under shortest-path routing, algorithmic betweenness scales as k^ν, with ν usually between 1.1 and 1.3.
- Bounded Delivery Rate: For η = 0.8, epidemic incidence remains significantly below the unbounded-delivery case across traffic and spreading rates.The comparison is reported for shortest-path delivery on random scale-free networks.
- Bounded Delivery Rate: With finite delivery capacity, congestion bounds epidemic incidence because additional injected packets do not increase the average packet flow through the network.The average infected population consequently remains roughly constant as more packets are injected.
- Bounded Delivery Rate: When η < ν, congestion arises; when η > ν, the epidemic-threshold diagram matches the unbounded-traffic case.The plateau of βc marks stationary global congestion, while congestion begins where the power-law dependence breaks down.
III. CONCLUSIONS
The paper develops analytical and numerical conditions for traffic-driven outbreaks in scale-free networks. It finds that flow determines epidemic thresholds and incidence, while finite delivery capacity introduces congestion-related limits.
- III. CONCLUSIONS: Analytical and numerical results show that epidemic thresholds are determined by contact flows and traffic-defined propagation pathways.The study examines disease contagion driven by traffic or interaction flow in scale-free networks.
- III. CONCLUSIONS: With sufficiently large or unbounded delivery capacity, the epidemic threshold is inversely proportional to traffic flow and vanishes at high traffic even in finite networks.The relation also depends on graph topology through algorithmic betweenness and recovers reaction-diffusion results as a particular case.
- III. CONCLUSIONS: The framework connects traffic-driven epidemic spreading to possible applications including air-transportation disease propagation and traffic-flow control.The paper discusses seasonal flow changes, air-traffic restrictions, and quarantining sensitive traffic nodes as related applications.
- III. CONCLUSIONS: The approximation lacks the detail needed to assess more realistic epidemiological models, but it is presented as a framework for anticipating general features of detailed agent-based models.This limitation qualifies the framework's intended level of epidemiological realism.
A. Structure
The study constructs scale-free networks using hidden geometric coordinates and power-law expected degrees. This structure can represent random or clustered small-world networks with properties found in real systems.
- A. Structure: Nodes are assigned uniformly distributed polar angles and expected degrees drawn from a distribution before connections are formed using hidden coordinates.The hidden metric space is represented by a circle with node coordinates combining angular position and expected degree.
- A. Structure: Choosing x(k) as a power-law distribution generates random networks with degree exponent γ > 2.The construction uses k > k0 and an average degree ⟨k⟩ to set the distribution's lower bound.
B. Routing Dynamics
Packets are generated continuously with random origins and destinations and routed through the network. The routing dynamics provide the traffic interactions that drive epidemic transmission.
- B. Routing Dynamics: At each time step, p = λN non-interacting packets receive randomly chosen origins and destinations, then are delivered using greedy routing.The model omits queues in this routing description.
C. Epidemic Dynamics
The SIS model represents nodes as healthy or infected, with infection transmitted through packet exchanges and recovery occurring at a fixed rate.
- Nodes occupy two states: healthy (S) or infected (I).
- Infection begins from an initial infected fraction ρ0 = I0/N.
- A susceptible node becomes infected with probability β when receiving a packet from an infected neighbor.
- Infected nodes recover at rate µ, fixed to 1 for most simulations.