Research on the Construction Method of Kill Chain Based on Deep Reinforcement Learning

The reasonable selection of operational units to form a kill chain is of great significance in modern warfare. To address this issue, this paper, based on the F2T2EA process, studies the modeling of heterogeneous nodes in the operational network, establishes a multi-objective effectiveness evaluation model for the reconnaissance and strike links in the operational process, and proposes a kill chain construction method based on deep reinforcement learning. Firstly, various targets on the battlefield are abstracted into fixed-type nodes, and the kill chain is modeled according to the complete F2T2EA process. Secondly, according to the differences between typical reconnaissance platforms and strike platforms, a multi-objective kill chain effectiveness evaluation model including the success rate and time of kill chain closure is established. Moreover, a deep reinforcement model based on the kill chain construction problem is built, and the PPO deep reinforcement learning algorithm is used to train the agent to solve the kill chain construction problem. The simulation results show that in the process of optimal kill chain construction, the solution obtained by the deep reinforcement learning algorithm is better than the approximate optimization algorithm widely used in engineering, and has certain engineering application value.