The Attack and Defense of Weakest-Link Networks

We experimentally test the qualitatively different equilibrium predictions of two theoretical models of attack and defense of a weakest-link network of targets. In such a network, the attacker’s objective is to successfully attack at least one target and the defender’s objective is to defend all targets. The models differ in how the conflict at each target is modeled — specifically, the lottery and auction contest success functions (CSFs). Consistent with equilibrium in the auction CSF model, attackers utilize a stochastic “guerrilla-warfare” strategy, which involves randomly attacking at most one target with a random level of force. Inconsistent with equilibrium in the lottery CSF model, attackers use the “guerrilla-warfare” strategy and attack only one target instead of the equilibrium “complete-coverage” strategy that attacks all targets. Consistent with equilibrium in both models, as the attacker’s valuation increases, the average resource expenditure, the probability of winning, and the average payoff increase (decrease) for the attacker (defender).