Controle de sistemas ciberfísicos baseado em defesa de alvo móvel

Abstract

This work proposes and validates a cybersecurity architecture based on Moving Target Defense (MTD) for the protection of cyber-physical systems (cps), focusing on the voltage regulation of a DC-DC Boost converter. Modern control systems, while highly efficient, are inherently vulnerable to stealthy False Data Injection (FDI) attacks, which can corrupt sensor readings and manipulate the physical control effort without triggering conventional alarms. To mitigate this threat, the developed architecture integrates a proactive and a reactive defense layer. The proactive layer promotes stochastic switching among a bank of optimal controllers (LQR with integral action) and state observers, dynamically altering the gain matrices and active sensors (voltage and current). Global stability of the switched system was ensured by calculating the minimum dwell-time using Linear Matrix Inequalities (LMI) based on Lyapunov Stability Theory. The reactive layer, in turn, employs the Bellman Error Integral to monitor discrepancies in the equations of the state observer. Simulation results obtained in a Python computational environment prove that, under a continuous FDI attack, the baseline plant without MTD suffers severe overvoltage due to mathematical estimation divergence, whereas the activation of the MTD architecture invalidates the attacker's static learning, detecting the anomalous data injection and limiting the physical impact to acceptable voltage oscillations, guaranteeing the integrity and resilience of the closed-loop converter.