A Framework for Joint Attack Detection and Control Under False Data Injection
A Framework for Joint Attack Detection and Control Under False Data Injection
复制标题
DOI：
10.1007/978-3-030-32430-8_21
复制
发表时间：
2019-10
期刊：
影响因子：
--
通讯作者：
Luyao Niu;Andrew Clark
中科院分区：
文献类型：
--
作者：
Luyao Niu;Andrew Clark
在这项工作中，我们考虑了在虚假数据注入攻击下具有卡尔曼滤波器、检测器和线性二次型高斯(LQG)控制器的LTI系统。控制器和对手之间的相互作用被一个Stackelberg博弈所捕获，其中控制器是领导者，对手是跟随者。我们提出了一种框架，在该框架下，系统选择时变的检测阈值来降低攻击的有效性，提高控制性能。我们将探测器的撞击建模为切换信号，从而得到切换的线性系统。首先利用所提出的框架计算最优攻击的闭合形式解，作为对任何检测阈值的最佳响应。然后，我们提出了一个凸规划来计算最优检测门限。我们的方法通过一个数值案例研究进行了评估。
In this work, we consider an LTI system with a Kalman filter, detector, and Linear Quadratic Gaussian (LQG) controller under false data injection attack. The interaction between the controller and adversary is captured by a Stackelberg game, in which the controller is the leader and the adversary is the follower. We propose a framework under which the system chooses time-varying detection thresholds to reduce the effectiveness of the attack and enhance the control performance. We model the impact of the detector as a switching signal, resulting in a switched linear system. A closed form solution for the optimal attack is first computed using the proposed framework, as the best response to any detection threshold. We then present a convex program to compute the optimal detection threshold. Our approach is evaluated using a numerical case study.