From Protection to Fragility: Emergent Failure Modes in Automated Cyber–Physical Systems

Highly automated cyber–physical infrastructures increasingly rely on coordinated defensive policies to contain attacks and maintain service continuity. While such coordination is commonly assumed to enhance resilience, its system-level consequences remain poorly understood when attackers adapt and defensive actions become predictable. This paper develops a stochastic modeling framework based on continuous-time Markov chains to examine how interacting attack and defense processes unfold in a tightly coupled automated building. The model represents progressive degradation, recovery, and irreversible global compromise, and characterizes security through the full distribution of Time-to-Compromise rather than through mean values alone. Survival functions, tail probabilities, and high-quantile risk measures are derived analytically from the generator of the chain. Through an extensive computational study, we identify parameter regimes in which stronger defensive alignment shortens rather than lengthens the expected lifetime of the system by inducing synchronization between adversarial and protective dynamics. Disruptive effects and delayed policy switching can partially counteract this mechanism, but often only after prolonged transient phases. These results suggest that resilience in automated environments is not determined solely by local defensive optimality, but by how interaction patterns evolve over time.