Optimization under attack: Resilience, vulnerability, and the path to collapse

Optimization is critical for improving the operations of large-scale socio-technical infrastructures such as those found in energy, mobility, and information systems. In particular, understanding the performance of multi-agent discrete-choice combinatorial optimization under distributed adversarial attacks is a compelling and underexplored problem. Multi-agent systems involve a large number of remote control variables that can influence the cost-effectiveness of distributed optimization heuristics. This paper unravels, for the first time, the trajectories of distributed optimization from resilience to vulnerability, and finally to collapse under varying adversarial influence. Using real-world and synthetic data to generate over 112 million multi-agent optimization scenarios, we systematically assess how the number of agents with varying levels of adversarial severity and network positioning influences optimization performance, with particular attention to the impact on Pareto optimality. With this large-scale dataset, made openly available as a benchmark, we disentangle how optimization systems remain resilient to adversaries and which adversary conditions make optimization vulnerable or cause collapse. These findings can support the design of self-healing strategies for fault tolerance and fault correction, addressing a critical gap in adversarial distributed optimization.