为揭示信息物理系统（CPS）潜在的安全风险并牵引出更先进的CPS安全技术，隐蔽攻击方法研究近年来备受关注。已有的隐蔽攻击方案大部分都是在攻击者已知CPS模型参数的假设下给出的；仅有的几种不依赖模型参数的隐蔽攻击方法需要假定攻击者已知CPS传感器测量值，而且它们的适用对象都是无给定信号CPS。本项目旨在移除以上两条保守假设，进而给出更一般的CPS（即具有给定信号闭环CPS）的隐蔽攻击策略。为此，在CPS模型参数未知且存在控制延迟的条件下，首先基于数据驱动技术建立隐蔽攻击技术验证平台（即基于TCP/IP的网络化交流伺服位置控制系统）；其次，针对分别由集线器和交换机组建的CPS，深入探讨实现信息侦听的技术途径，进而探索阻断网络通信以及插入伪造数据的方法；然后，论证CPS动态特性矩阵的数据驱动闭环迭代辨识方法；最后，利用以上研究结果给出数据驱动隐蔽攻击方案，并通过实验核实该方案的有效性和实用性。

To reveal the potential security risks of cyber-physical systems (CPS) and develop more advanced security technologies for CPS, the study of stealthy attack methods has gotten more attention in recent years. Most of the existing stealthy attack schemes are presented under the precondition that the model parameters of CPS are known to attackers. At present, there are only a few model-independent stealthy attack approaches, which are applicable only to the CPS without reference signals and need the assumption that attackers know sensor readings. This project aims to remove the aforementioned two conservative assumptions and further give a stealthy attack methodology for more general CPS, i.e., the closed-loop CPS with reference signals. To this end, under the condition that the model parameters of CPS are unknown and such systems have control delays, a platform (i.e. a TCP/IP-based networked AC servo position control system) used for the verification of the stealthy attack technology is firstly constructed via data-driven methods. Then, a plan is made, which is utilized for eavesdropping the information of the CPS built by employing, respectively, Ethernet hubs and switches. On this basis, an approach to blocking network communications and injecting the false sensor data into CPS is explored. After that, according to data-driven methods, a closed-loop recursive identification strategy for the dynamic characteristic matrix of CPS is designed. Finally, with the aid of all the above-obtained results, the data-driven stealthy attack scheme for the CPS is proposed, and subsequently, its effectiveness and practicability are validated by experiment.