Abstract: In order to facilitate the research and development of the security defense system for large pressurized heavy water reactors (PHWR), this paper studies the potential attack mode in PHWR networked control system and proposes a covert attack approach based on Gaussian process regression model optimized by salp swarm algorithm. In this method, when the false data is injected into the PHWR networked control system, the system identification is addressed by optimizing the Gaussian process regression algorithm and a high-precision estimation model of the attacked area in PHWR is obtained, and then the estimation model is used to realize the covert attack. The simulation results show that the attack method not only causes some damage to PHWR, but also has high concealment performance.