Attacks detection in direct data-driven control

The basic scope of the study of this thesis work is attack detection in the direct data-driven control. Attacks which are potentially may violate the work of sensors and actuators, accordingly an outcome of the controller. The main concept of the direct data-driven control is to design a controller, which will correspond output of the system to a given reference model, having in disposition only measured input and output data, in other words without knowing the exact structure of the plant itself. Taking into consideration the errors that may interfere with the signal during the measurement of the input and output data, the feasible controller parameter set is defined. Then the controller is designed by choosing suitable parameters from the previously defined feasible set. In order to identify false control commands to actuators and false sensor readings, in this paper, the new approach of attack detection will be studied. The study is divided into three main steps, which are: 1. Detection of an attack in a closed feedback loop system. 2. Distinguishing the attacks on the sensor and actuator separately. 3. Possible definition of the set of controller parameters in order to control distortion. The study of the above-mentioned points is based on simulations triggered by two types of attack injection: 1. Attacks considered as an additive signal. 2. Attacks changed the parameter values of sensors and/or actuators.