Regret-Optimal Defense Against Stealthy Adversaries: A System Level Approach
Modern control designs in robotics, aerospace, and cyber-physical systems rely heavily on real-world data obtained through system outputs. However, these outputs can be compromised by system faults and malicious attacks, distorting critical system information needed for secure and reliable operation. In this paper, we introduce a novel regret-optimal control framework for designing controllers that make a linear system robust against stealthy attacks, including both sensor and actuator attacks. Specifically, we present (a) a convex optimization-based system metric to quantify the regret under the worst-case stealthy attack (the difference between actual performance and optimal performance with hindsight of the attack), which adapts and improves upon the $\mathcal{H}_2$ and $\mathcal{H}_{\infty}$ norms in the presence of stealthy adversaries, (b) an optimization problem for minimizing the regret of (a) in system-level parameterization, enabling localized and distributed implementation in large-scale systems, and (c) a rank-constrained optimization problem equivalent to the optimization of (b), which can be solved using convex rank minimization methods. We also present numerical simulations that demonstrate the effectiveness of our proposed framework.
Code (0)
등록된 구현이 없습니다.
Similar Papers 제목 키워드 기반
DROP: Poison Dilution via Knowledge Distillation for Federated Learning
Federated Learning is vulnerable to adversarial manipulation, where malicious clients can inject poisoned updates to influence the global model's behavior. While existing defense mechanisms have made notable progress, th…
Data PoisoningFederated LearningKnowledge DistillationStealthy Poisoning Attacks Bypass Defenses in Regression Settings
Regression models are widely used in industrial processes, engineering, and in natural and physical sciences, yet their robustness to poisoning has received less attention. When it has, studies often assume unrealistic t…
Active Defense Against False Data Injection Attacks in Robotic Manipulators
Robotic systems are vulnerable to False Data Injection Attacks (FDIAs), where adversaries corrupt sensor signals to gain malicious control. Feedback linearization exposes robotic systems to integrator vulnerability, maki…
Avoid Adversarial Adaption in Federated Learning by Multi-Metric Investigations
Federated Learning (FL) facilitates decentralized machine learning model training, preserving data privacy, lowering communication costs, and boosting model performance through diversified data sources. Yet, FL faces vul…
Federated LearningHierarchical Multi-agent Reinforcement Learning for Cyber Network Defense
Recent advances in multi-agent reinforcement learning (MARL) have created opportunities to solve complex real-world tasks. Cybersecurity is a notable application area, where defending networks against sophisticated adver…
Multi-agent Reinforcement Learningreinforcement-learningReinforcement Learning