Control Barrier Function-Based Attack Recovery With Provable Guarantees

• 成果类型: Article
• 署名作者: Garg, Kunal; Sanfelice, Ricardo G.; Cardenas, Alvaro A.
• 署名单位: Arizona State University; Arizona State University-Tempe; University of California System; University of California Santa Cruz; University of California System; University of California Santa Cruz
• 刊物名称: IEEE TRANSACTIONS ON AUTOMATIC CONTROL
• ISSN/ISSBN: 0018-9286
• DOI: 10.1109/TAC.2026.3683371
• 发表日期: 2026
• 关键词: systems; Invariance; sets

摘要

This article investigates security guarantees for cyber-physical systems (CPS) against actuator attacks. We introduce a new attack-detection mechanism based on zeroing control barrier function conditions. We propose an adaptive recovery mechanism that responds based on the system's proximity to safety violations. Our attack-detection mechanism has been proven to be sound, meaning that it consistently detects adversarial attacks without any false negatives. In addition, we propose a novel hybrid control law that addresses delays in attack detection and prevents Zeno behavior. We also propose a sampling-based method to verify whether a set is a viability domain for CPS. Finally, we employ a quadratic programming approach for synthesizing control laws for the hybrid control policy, utilizing the viability domain to ensure safety in the presence of adversarial attacks on system actuators. The efficacy of the proposed method is demonstrated in a simulation case study involving a quadrotor system.