Modeling the impact of disinformation on infrastructure networks: A node-to-path causal failure approach across interconnected layers

The widespread dissemination of disinformation poses a serious threat to society, particularly if such disinformation is weaponized to target critical infrastructure. By conceptualizing the spread of disinformation as a dynamic process on one network layer (denoted as A ) and the physical infrastructure network as another (denoted as B ), we can observe the indirect impacts of the failure of the components in A on the components in B . For example, false messages about potential power rate hikes or outages (in A ) can manipulate consumer behavior, resulting in increased power demand that might overload transmission lines, ultimately leading to blackouts (manifesting as disruptions in B ). Such disruptions highlight the interconnected vulnerabilities between disinformation dissemination and critical infrastructure. Unlike previous studies that broadly examine interdependencies between networks, our approach captures direct cause-and-effect relationships. We model these relationships across layers using the concept of causal failures. Specifically, we formally define causal failures across both layers and introduce the concept of stochastic causal failure, where the probability of a node-to-path failure relationship is influenced by stochastic processes. We then propose a problem to study these stochastic causalities, focusing on how the failure of a node in layer A stochastically impacts the failure of a path in layer B . To address this, we develop an algorithm to determine the maximum number of stochastic causalities in A needed to create a giant component of a specified size in B . Furthermore, we present another algorithm that calculates the minimum number of stochastic causal failures in A required to generate at least k small connected components in B . Our experiments examine how the spread of disinformation in layer A can activate causalities, leading to stochastic failures in the infrastructure layer B , which we model with data describing social media and power networks.