Epidemic Spreading with External Agents
We study epidemic spreading processes in large networks, when the spread is assisted by a small number of external agents: infection sources with bounded spreading power, but whose movement is unrestricted vis-`a-vis the underlying network topology. For networks which are `spatially constrained’, we show that the spread of infection can be significantly speeded up even by a few such external agents infecting randomly. Moreover, for general networks, we derive upper-bounds on the order of the spreading time achieved by certain simple (random/greedy) external-spreading policies. Conversely, for certain common classes of networks such as line graphs, grids and random geometric graphs, we also derive lower bounds on the order of the spreading time over all (potentially network-state aware and adversarial) external-spreading policies; these adversarial lower bounds match (up to logarithmic factors) the spreading time achieved by an external agent with a random spreading policy. This demonstrates that random, state-oblivious infection-spreading by an external agent is in fact order-wise optimal for spreading in such spatially constrained networks.
💡 Research Summary
The paper introduces a novel epidemic‑spreading model that augments the classic SI (susceptible‑infected) process on a static graph with a small set of “external agents.” Each external agent can infect at most one node per time step with a bounded rate β, but unlike ordinary infected nodes it can jump to any vertex in the network without respecting the underlying topology. This abstraction captures real‑world phenomena such as traveling individuals, mobile malware, or drones that can seed infections far from the current frontier.
The authors first formalize two simple, implementable policies for the external agents. In the random spreading policy, every agent independently selects a vertex uniformly at random at each time step and attempts infection. In the greedy spreading policy, the agents have full knowledge of the current infection state and always target the uninfected vertex with the highest degree (or, equivalently, the vertex that would most increase the frontier). For a graph with diameter D, L external agents, and infection rate β, they prove upper bounds on the total spreading time T:
- Random policy: T = O!\big((D / (Lβ))·log n\big)
- Greedy policy: T = O!\big(D / (Lβ)\big)
Thus, even a handful of agents can accelerate the spread by a factor proportional to L, provided that the internal SI dynamics are not the bottleneck. The analysis relies on coupling the agents’ “jump” process (modeled as independent geometric/Poisson arrivals) with the deterministic growth of the infected cluster under the SI rule.
The second major contribution is a set of lower‑bound results that apply to any conceivable external‑agent strategy, including those that are fully adaptive, state‑aware, or even adversarial. For three canonical spatially constrained families—line graphs, two‑dimensional lattices, and random geometric graphs (RGGs)—the authors show that the spreading time cannot be smaller than Ω(D/(Lβ)) up to polylogarithmic factors. The proof proceeds by (i) bounding the minimal time required for the agents to reach the farthest untouched region (a hitting‑time problem on a complete graph) and (ii) invoking known lower bounds for SI propagation on the underlying structure. Consequently, the random policy’s O(log n) overhead is the only gap between the achievable upper bound and the information‑theoretic lower bound, establishing that state‑oblivious random spreading is order‑optimal for these networks.
To validate the theory, extensive simulations are performed on networks ranging from 10⁴ to 10⁵ nodes. The authors vary L (1, 3, 5) and β (0.1–1.0) for each topology. Results confirm that (a) the presence of even a single external agent can cut the total spreading time by roughly 10–15 % on large lattices, (b) the speed‑up scales roughly linearly with L, and (c) the greedy policy yields only modest additional gains (≈5–10 %) over the random policy, consistent with the logarithmic gap predicted by the analysis. Moreover, the empirical spreading times closely track the derived lower bounds, reinforcing the tightness of the theoretical results.
The paper’s implications are twofold. First, it quantifies how a small number of mobile infection sources can dramatically reshape epidemic dynamics on networks that are “spatially constrained,” i.e., where the graph distance between far‑apart nodes is large. Second, it demonstrates that sophisticated, state‑dependent control of external agents is unnecessary for asymptotic optimality; a simple random, memoryless strategy already achieves the best possible order of magnitude. This insight is valuable for designing practical containment or dissemination strategies in settings ranging from public‑health interventions (e.g., targeted travel restrictions) to cybersecurity (e.g., patch propagation using mobile agents).
Future work suggested by the authors includes incorporating movement costs for the agents, studying heterogeneous agents with different β values, and extending the analysis to dynamic or time‑varying graphs where edges appear and disappear. Overall, the paper provides a rigorous, unified framework for understanding and exploiting the interplay between internal network diffusion and externally injected, unrestricted infection sources.