Cyber-Physical Systems (CPS) are systems that include physical and computational 
components linked by communication channels. In a Smart Grid (SG), the power plants and loads 
communicate with supervisors (Central Controllers (CC)) for managing the power demand more 
efficiently. As such, a smart grid can be regarded as a CPS. The computational components and 
communication links of a CPS can be subject to cyber-attacks. Researchers have been exploring 
detection and mitigation strategies for various types of cyber-attacks. 
An important type of attack is the replay attack for which various strategies based on 
watermarking signals have been proposed. One such scheme is based on switching multi-sine 
waves as the watermarking signal. This thesis adapts this scheme and develops a design procedure 
for detecting replay attacks for smart grids. Specifically, it examines the places in a grid where the 
watermarking signal can be injected and presents guidelines for choosing the amplitude and 
frequencies of sine waves that suit smart grids. 
One of the drawbacks of using a watermarking signal is the additional control cost (i.e., 
decrease in performance). In the context of smart grids, watermarking results in small fluctuations 
in delivered power. This thesis extends the single-input-single-output watermarking to a two-input-two-output watermarking scheme for smart grids in such a way to considerably lower grid power 
fluctuations due to watermarking. The proposed method is verified using a simulated grid 
connected inverter-based plants. Simulation results show that using the suggested strategy, the 
effect of watermarking on the overall grid power reduces significantly.