Destabilizing dynamic networks under conditions of uncertainty

Managing and controlling knowledge intensive dynamic systems requires being able to estimate, analyze, and evaluate, under conditions of uncertainty, the existing system and the impact of actions, such as changes in personnel or resources on this system. Our approach to this problem is dynamic network analysis. Using a combination of statistical and simulation tools, we analyze the robustness under uncertainty of a series of metrics for identifying key entities whose removal from the network destabilizes the network by degrading performance on one or more dimensions. We examine multiple types of uncertainty, including cases of over and underestimation of the presence of relations among entities. We find that higher levels of information assurance are needed for nodes than edges, and that destabilization even under uncertainty can still be disruptive. A potential application is to identify individuals in a covert network to isolate in order to decrease the network's ability to take action.