The invention discloses a physical layer security resource allocation method in an ICV network, solves the problem that an existing allocation mode cannot optimize V2V time delay, and belongs to the technical field of wireless communication. The method comprises the following steps: S1, a sending vehicle in each V2V link being an intelligent agent, the intelligent agent being realized by adoptinga deep reinforcement learning network and comprising a trained target network and a real-time network, and the target network being used for updating parameters of the real-time network; S2, each intelligent agent obtaining local observation of the environment, including information of legality, interference and eavesdropping channels, sending load, time delay limitation, training iteration timesof a target network and exploration utilization probability; S3, each intelligent agent outputting an action by using a real-time network according to local observation, wherein the action comprises afrequency band occupied by signal sending and transmitting power; S4, the target networks of all the intelligent agents receiving the same reward from the environment, the environment entering the global state of the next moment, executing the step S2, and the reward comprising the system secrecy capacity and the time delay limit.