An Adaptive Secure Control Scheme for T-S Fuzzy Systems Against Simultaneous Stealthy Sensor and Actuator Attacks
An Adaptive Secure Control Scheme for T-S Fuzzy Systems Against Simultaneous Stealthy Sensor and Actuator Attacks
复制标题
针对同时隐形传感器和执行器攻击的 T-S 模糊系统的自适应安全控制方案
DOI：
10.1109/tfuzz.2020.2990772
复制
发表时间：
2021
影响因子：
11.9
通讯作者：
Dong Jiuxiang
中科院分区：
文献类型：
--
作者：
Huang Xin;Dong Jiuxiang
研究了T-S模糊系统在同时受到传感器和执行器攻击时的安全控制问题。一个关键的问题是，隐形攻击可以恶意妥协的控制输入，传感器输出，和模糊控制器的前提变量取决于可测量的状态，而绕过检测从异常检测器，使他们可能离开传统的控制方案无效。针对这一问题，1）考虑到传感器攻击对模糊控制器前提变量的影响，提出了一种由可信传感器和引入滤波器的输出构成可靠前提变量的模糊控制器，并利用模糊乘积推理机的性质给出了保证期望H ∞性能的控制器设计条件。2)借助检测机制和攻击者的结构信息，给出了一个隐身攻击的上界，从防御者的角度量化了一类隐身攻击。在此基础上，提出了一种新的自适应攻击补偿器，使得具有该补偿器的模糊控制器能够保证被攻击系统的稳定性，并具有期望的控制性能。三个示例验证了所提出的方法。
This article studies secure control problems for T-S fuzzy systems equipped with a Luenberger-observer-based anomaly detector, when simultaneously subjected to stealthy sensor and actuator attacks. A key issue is that stealthy attacks can maliciously compromise control inputs, sensor outputs, and fuzzy controller's premise variables which depend on measurable states, while bypassing detection from an anomaly detector, so that they may leave the conventional control schemes invalid. To solve it, 1) by considering the impact of sensor attacks on premise variables of fuzzy controller, a new fuzzy controller with reliable premise variables constructed by the trusted sensor's and introduced filter's outputs is proposed and also a design condition for the controller with the guaranteed desired H ∞ performance is given via the property of fuzzy product inference engines. 2) With the aid of the detection mechanism and attack's structural information, an upper bound for stealthy attacks is given, which from the defender's viewpoint quantifies a class of stealthy attacks. Based on the bound, a novel adaptive attack compensator is presented such that the developed fuzzy controller with the compensator guarantees the stability of the attacked systems with a desired control performance. Three illustrative examples validate the proposed method.