Network traffic control method of NHP based on deep reinforcement learning

The Network-Infrastructure Hiding Protocol (NHP) represents a secure communication protocol that substantially reduces attack surfaces by concealing network resources by default and granting minimal access privileges on-demand following identity and authorization verification. However, optimizing Quality of Service (QoS) network traffic control within NHP environments remains a critical challenge. Although Software-Defined Networking (SDN) offers potential solutions through centralized control and dynamic resource allocation, current scheduling methods based on Deep Reinforcement Learning (DRL) exhibit limitations including constrained optimization effectiveness and insufficient dynamic adaptability. This paper proposes an intelligent regulation method based on DRL to address network traffic control challenges in NHP network environments. This paper introduces the Dueling Double Deep Q-Network (D3QN) algorithm to construct an agent system capable of real-time network state perception and autonomous decision-making. Experimental results demonstrate that the proposed method significantly outperforms traditional algorithms across key indicators including throughput, latency, and packet loss rate, exhibiting exceptional adaptability and stability particularly in dynamic network environments. This research provides an efficient and reliable intelligent control solution for traffic optimization in complex network environments, offering novel theoretical and practical pathways for addressing core challenges in performance-security collaboration within next-generation network management and control systems, with strong practical value and promising application prospects.