This paper addresses the filtering problem for a class of discrete-time cyber-physical system (CPS) in simultaneous presence of physical attack, denial of service attacks (DoS), false data injection (FDI) attacks and randomly occurring nonlinearities (RONs). The malicious attacks can be simultaneously launched at the physical and cyber space to assault/disturb the system's state or block the measurement transmission channels. Firstly, the physical attack is considered as an injected unknown information, which can modify the system's state; then, a unified framework is proposed under which the phenomena of DoS attacks and FDI attacks are depicted by two independent Bernoulli distributed random variables, respectively. Furthermore, the available random probability information of the hybrid attacks and RONs are taken into account in the developed state estimation algorithm. Subsequently, by employing the stochastic analysis and matrix techniques, the exponential boundedness of the state estimation error is discussed from the theoretical aspects. As a typical CPS, a quadrotor unmanned aerial vehicle(UAV) system is provided to demonstrate the effectiveness and practicability of the proposed state estimation method.

本文解决了同时存在物理攻击、拒绝服务攻击 (DoS)、虚假数据注入 (FDI) 攻击和随机发生的非线性 (RON) 的一类离散时间网络物理系统 (CPS) 的过滤问题。恶意攻击可以同时在物理空间和网络空间发起，以攻击/扰乱系统状态或阻塞测量传输通道。首先，物理攻击被认为是一种注入的未知信息，可以修改系统的状态；然后，提出了一个统一的框架，在该框架下，DoS 攻击和 FDI 攻击的现象分别用两个独立的伯努利分布随机变量来描述。此外，在开发的状态估计算法中考虑了混合攻击和 RON 的可用随机概率信息。随后，通过使用随机分析和矩阵技术，从理论方面讨论了状态估计误差的指数有界性。作为典型的CPS，提供了一个四旋翼无人机（UAV）系统来证明所提出的状态估计方法的有效性和实用性。