Assessing the robustness of physical networks under attack uncertainty

Network robustness is fundamental to understanding and mitigating the vulnerabilities of complex physical infrastructure networks. Traditional studies often evaluate robustness by examining network performance under various attack strategies, typically assuming that attacks are always successfully implemented. However, physical infrastructure attacks are inherently uncertain, and their success is not guaranteed due to defensive mechanisms, resource limitations, or strategic failures. To address this gap, we propose a novel network attack uncertainty model that explicitly accounts for attack failures, providing a more realistic framework for robustness evaluation. Unlike existing studies that assume deterministic attack implementation, our model captures the uncertainty nature of physical infrastructure disruptions targeting nodes or edges. Using this model, we conduct extensive experiments on both synthetic networks—including Barabási–Albert (BA), Watts–Strogatz (WS), and Erdős–Rényi (ER) models and real-world international trade networks. Our findings challenge the conventional wisdom that targeted attacks on central nodes are the most effective. Surprisingly, when accounting for attack uncertainty, peripheral attacks inflict the greatest damage on network robustness, followed by random attacks, while central attacks have the least impact. This unexpected result underscores the vulnerability of peripheral structures, which are often overlooked in traditional robustness assessments. These insights offer new perspectives on network defense strategies, highlighting the need to protect not only central nodes but also peripheral components that may serve as critical points of failure under uncertain attack conditions. Our work provides a more accurate robustness assessment framework for physical networks and contributes to the development of more resilient network defense strategies against unpredictable threats.