Cyber-physical systems (CPS), serving as the cornerstone of Industry 4.0 and smart grids, feature deep integration between cyber and physical layers. While this integration enables high efficiency, it simultaneously introduces significant security vulnerabilities. Cyberattacks pose severe threats to CPS operations, potentially causing physical damage and system failures. This paper systematically reviews recent advances in CPS security from a defensive perspective, with focus on three key technologies: secure state estimation, security control, and attack detection. First, we categorize typical attack models by attack type and location. Next, the state-of-the-art in secure state estimation are summarized. To address sensor data tampering, researchers have developed methods based on exhaustive search, convex optimization, and artificial intelligence to reconstruct system states under sparse attacks. Then, security control strategies against cyber-layer attacks and collision/obstacle avoidance methods for physical-layer disturbances are examined. Subsequently, model-based and data-driven attack detection approaches are systematically summarized. Finally, the paper summarizes the current limitations and challenges in the field and outlines potential avenues for future research.