Research On Security Theory For Cyber-Physical Systems

With the rapid progress of information technology, human society has gradually stepped into the information era, which is characterized by digitization, networking and intelligence. Informa-tization systems, including intelligent transportation, smart factory, smart grid, electronic finance, social network, have sprung up and dramatically changed the mode of life, production and man-agement of human society. These systems utilize information sensing, ubiquitous computing, and management to interconnect between physical space, cyber space, and human society, and then move towards deeper integration. Eventually cyber-physical systems are birthed.The interconnected future world necessarily requires higher-level openness of CPS, which will inevitably result in higher potential security threats. Statistical data on security incidents shows that the security threats are increasing exponentially and the consequences are becoming increasingly serious. Therefore, security issues of CPS have aroused concerns of governments and social sectors around the world, and the related researches are tremendously supported.CPS are the networked systems with feedback mechanisms essentially. Thus, it is of both practical and theoretical significance to study the CPS security from the perspective of control. The corresponding researches include attack modelling, security issues on state estimation and control performances, privacy preserving problems, and so on. Although great progress has been made in this field, there still exist many shortcomings. For instance, the current research on the se-curity of state estimation and control performance mainly focus on the robust or resilient estimators and controllers, but it is neglected to study the optimal attack policy with energy constraint. For the privacy preserving issues, most researches put emphasis on the design of privacy preserving mechanisms, and the balance between privacy preserving and control performance is neglected.In view of these shortcomings, this doctoral dissertation studies the optimal attack policy against remote state estimation, online deception attack strategy against state estimation, optimal attack scheduling against LQG control, and the trade-off between privacy preserving and system performance. The specific research contents in this dissertation are as follows:1. An overview of existing literature on CPS security is provided, and deficiencies of this re-search field have been analyzed.2. For the problem of energy limited DoS attack against remote estimation, the optimal DoS attack scheduling schemes are proposed. For two typical system performance indexes, i.e., the expected average estimation error and the expected terminal estimation error, the optimal attack schedules are constructed, respectively. The optimal attack schedules against special attack detection mechanism is also provided. In addition, the characteristic of optimal attack schedules is given when the sensor has energy constraint as well.3. For the problem of deception attack against remote estimation, an online attack policy is pro-posed. In the viewpoint of attacker, it is investigated how to design proper deception attack strategy to degrade the state estimation quality with communication rate constraint. By the-oretical derivation, an online attack strategy is provided and the proof has demonstrated that the proposed attack strategy can degrade the estimation quality. To study the effectiveness of the proposed strategy, the cost deviation, which depicts the difference between the estimation quality with and without the proposed attack strategy, is analyzed.4. For the energy limited DoS attack against optimal control, optimal DoS scheduling schemes are provided. The optimal DoS attack schedules are designed to maximize the linear quadratic Gaussian control cost function under energy constraint. After the properties of the cost func-tion under an arbitrary attack schedule are analyzed, the optimal attack schedules and the corresponding cost function are derived. System stability under this optimal attack sched-ules is then considered. The optimal attack schedule in CPS with multiple subsystems is also investigated.5. The tradeoff between individual privacy and system performance in CPS is considered. The performance optimization problem subject to the given differential privacy requirement is formulated. For a simplified system, the close-form expression of optimal system perfor-mance under desired privacy requirement is provided. The future research topics on privacy preserving problem of CPS are given in the end.In the end, the thesis is concluded and some future research works are discussed.