The security of modern power systems is a pressing concern due to their real-time requirements and the integration of various technologies. Recent cyber incidents have highlighted the limitations of traditional security measures based on information and communications technology (ICT). To address these challenges, this chapter makes significant contribution in the field of cybersecurity for smart grids. By developing innovative methods based on machine learning algorithms, dynamic analysis, and belief propagation techniques, this chapter aims to enhance the detection, prevention, and mitigation of cyber attacks in power systems. One key contribution of the chapter is the development of a methodology that utilizes machine learning algorithms for the detection of false data injection attacks (FDIAs) during power system state estimation. By leveraging the power of machine learning, this approach enhances the ability to identify and mitigate FDIAs effectively. Additionally, this chapter investigates the emergence of stealthy FDIAs, improving the understanding and detection capabilities of modern machine learning algorithms. Furthermore, this chapter emphasizes dynamic analysis over steady-state analysis, addressing the limitations of traditional approaches. By considering the dynamic behaviors of power systems, this chapter enhances the understanding and detection of cyber threats. This approach provides a more comprehensive assessment of system behavior, particularly in the context of cyber attacks, thereby strengthening the overall security of smart grids.