Print Email Facebook Twitter Watermarking for attack detection in networked control systems: comparison between a linear and a nonlinear approach Title Watermarking for attack detection in networked control systems: comparison between a linear and a nonlinear approach Author Mooren, Maurits (TU Delft Mechanical, Maritime and Materials Engineering; TU Delft Delft Center for Systems and Control) Contributor Ferrari, Riccardo M.G. (mentor) Degree granting institution Delft University of Technology Programme Mechanical Engineering | Systems and Control Date 2019-12-20 Abstract With the increasing amount of information sent in control systems, data is more often communicated through a communication network. These networked control systems use the network to exchange control and feedback signals among the system’s components. Data sent over a network attracts adversaries that try to read and modify this data. These attempts of modifying the data are unwanted and need to be detected. In order to detect these attacks a digital watermark is added to the data in real time. The digital watermark gives the opportunity to check whether the data is authentic. In order to do this, the watermark embedding function is data dependent where the watermark is created from old data points. The selection of which data point is used changes over time and the moment of changing these so called delay parameters is also data dependent. With this information alone an attack can not be detected. Steganography is used to create a signal independent attack detection technique that has a 100% detection rate for the investigated attacks. There are different digital watermarking techniques. Echo Hiding and Quantization Index Modulation (QIM) are a linear and nonlinear watermarking technique that have the possibility to be used in real-time. QIM gives a higher level of security while Echo Hiding can be implemented faster and takes less CPU load. QIM removes some of the accuracy of the data and Echo Hiding takes less memory. Both techniques have advantages and disadvantages which are compared and the two techniques are tested on two systems; a standard three-tank system where data (the fluid levels) change slowly over time and a much faster system, the Van der Pol oscillator. Both watermarking techniques detect the replay and reroute attack on both systems 100% of the testedtime. Subject networked control systemswatermarkingattack detection To reference this document use: http://resolver.tudelft.nl/uuid:51355f59-7a5e-4cd7-b708-45f4c8993386 Part of collection Student theses Document type master thesis Rights © 2019 Maurits Mooren Files PDF mscThesis_MMooren.pdf 8.3 MB Close viewer /islandora/object/uuid:51355f59-7a5e-4cd7-b708-45f4c8993386/datastream/OBJ/view