The cost of attack in competing networks

Real-world attacks can be interpreted as the result of competitive interactions between networks, ranging from predator-prey networks to networks of countries under economic sanctions. Although the purpose of an attack is to damage a target network, it also curtails the ability of the attacker, which must choose the duration and magnitude of an attack to avoid negative impacts on its own functioning. Nevertheless, despite the large number of studies on interconnected networks, the consequences of initiating an attack have never been studied. Here, we address this issue by introducing a model of network competition where a resilient network is willing to partially weaken its own resilience in order to more severely damage a less resilient competitor. The attacking network can take over the competitor nodes after their long inactivity. However, due to a feedback mechanism the takeovers weaken the resilience of the attacking network. We define a conservation law that relates the feedback mechanism to the resilience dynamics for two competing networks. Within this formalism, we determine the cost and optimal duration of an attack, allowing a network to evaluate the risk of initiating hostilities.

💡 Research Summary

The paper introduces a novel framework for quantifying the cost of initiating an attack in a pair of competing networks, addressing a gap in the literature where prior work has focused almost exclusively on the impact of attacks on the target network. The authors model two interacting networks—designated as the strong (S) and weak (W) networks—using either Barabási‑Albert (scale‑free) or Erdős‑Rényi (random) topologies, each endowed with intra‑network links and a set of inter‑network connections. Nodes can fail internally with probability p₁ and externally with probability p₂; external failure occurs when the fraction of active neighbors falls below a prescribed fractional threshold T (different for S and W, with T_S < T_W, making S more resilient).

The core of the model is the attacker’s ability to deliberately raise its own internal failure probability p₁, thereby increasing the failure rate in the opponent network. This act inevitably raises the attacker’s own failure rate, creating a feedback loop that reduces its resilience. Moreover, if a node in the weak network remains inactive for longer than a multiple n of the recovery time τ, the strong network can “take over” that node. The takeover incurs a cost: acquiring a node of degree k_{W,i} reduces the strong network’s resilience because the new node is intrinsically weaker. The authors formalize this trade‑off with a conservation law:

 N · h · k_{S,i}(T₀_S − T_S) = k_{W,i}(T_W − T₀_S),

where N is the size of S, h its average degree, T₀_S the post‑takeover threshold, and the right‑hand side quantifies the loss of resilience due to the acquired node.

Using mean‑field theory, the fractions of failed nodes a_S = 1 − f_S and a_W = 1 − f_W are expressed in terms of effective internal failure rates p*{S,1}=1−exp(−p{S,1}τ) and p*{W,1}, together with external failure terms E_S and E_W that depend on the degree distribution and the activation thresholds. Analytical results reveal a critical region: increasing p*{S,1} up to roughly 0.18 causes a sharp drop in the weak network’s active fraction f_W while the strong network still retains a high f_S. Beyond a second critical point around p*_{S,1}≈0.33, the strong network itself undergoes a first‑order transition and collapses. Consequently, an optimal attack must be confined to the interval where the incremental increase in failures of the weak network (Δa_W) exceeds that of the strong network (Δa_S). This condition corresponds to a region in the (a_S, a_W) plane where the slope of a_W versus a_S exceeds one.

Extensive simulations corroborate the analytical predictions. For BA networks with parameters m_S=m_W=3 and inter‑network links m_{S,W}=m_{W,S}=2, the authors demonstrate hysteresis: after raising p₁ to induce collapse of W, resetting p₁ to its original value does not restore W, whereas S recovers fully. Similar behavior is observed for ER networks and for assortatively mixed degree correlations, indicating that the phenomenon is robust to topology.

The paper also explores the strategic dimension of takeovers. The longer a weak node stays inactive (larger n), the higher the probability of acquisition, but each acquired node reduces S’s resilience proportionally to its degree. The conservation law quantifies this cost, allowing the attacker to compute an optimal combination of attack intensity (p*_{S,1}) and duration (n·τ) that maximizes net gain (damage inflicted on W) while keeping self‑damage below a tolerable threshold.

Practical implications are discussed across several domains: economic sanctions (where a powerful country’s trade restrictions harm both the target and the sanctioning country), military campaigns (resource depletion versus enemy degradation), and ecological predator‑prey dynamics (over‑exploitation of a prey species eventually weakening the predator). The framework provides a systematic way to evaluate “the cost of attack” and to determine the optimal timing and magnitude of hostile actions in any setting where competing networks interact.

In summary, the authors present a comprehensive theoretical and computational study of competitive network attacks, introducing a conservation‑law‑based cost metric, identifying critical thresholds for effective aggression, and outlining optimal attack strategies. This work fills a notable gap in network science by explicitly modeling the attacker’s self‑inflicted damage and offers a versatile tool for analyzing real‑world competitive interactions.