Date of Award
Spring 2019
Document Type
Thesis
Degree Name
Master of Science (MS)
Department
Electrical and Computer Engineering
First Advisor
Yaz, Edwin E.
Second Advisor
Schneider, Susan C.
Abstract
In this thesis, two different approaches to sensor intrusion detection are presented. In the first approach, an estimation algorithm using a bank of Kalman Filters is designed that is capable of estimating the intrusion signal when sensors are affected in control systems. The mathematical models of the control system will be established and the system measurement will be shown and after that, various false signals, such as constant-type and ramp-type signal, will be selected as the intrusion signal to affect the system output mentioned above. The system measurement will be tested based on a bank of Kalman Filters. The probabilities of each intrusion state (affected and unaffected) of the control system will be calculated as a function of time. The estimation of the states from a bank of Kalman Filters together with the associated probabilities will determine whether the sensor is under attack or not by using the information from the estimation algorithm. The performance of the algorithm will be tested based on the various levels of the system and measurement noise.In the second approach, a new estimation algorithm is applied to detect the intrusion signal targeting the system mentioned above. By calculating the sample mean value of the system state and measurement in time, the changes of the system measurement can be detected by calculating the residual between the actual value and the theoretical sample mean value of the system measurement and in that case, the intrusion signal can be found. Thesis conclusions, summary and future work is also mentioned in the last chapter of this work.