This talk addresses the problem of secure control of cyber-physical systems. We consider physical plants that are being controlled by multiple actuators and sensors, where some sensors could be subverted or “malicious.” A malicious sensor may not report the measurements that it observes truthfully. We address a general technique by which the actuators can detect the actions of malicious sensors in the system and disable closed-loop control based on their information. This technique, called “dynamic watermarking,” employs the technique of actuators injecting small private excitation into the system, which will reveal malicious tampering with signals. We establish that such an active defense can provide provable security guarantees for networked systems of sensors and actuators from arbitrary attacks.