GP-SR:基于GNN-PPO的低轨卫星网络安全路由优化算法OA

To address the routing optimization problem in low earth orbit(LEO)satellite networks under conditions of rapidly changing topology,locally observable states,and concurrent security attacks,a GNN-PPO based secure routing(GP-SR)algorithm was proposed,a risk-sensitive intelligent routing algorithm based on graph neural networks and proxi-mal policy optimization.The algorithm modelled distributed routing decision-making as a partially observable Markov decision process,leveraged history-enhanced observations to mitigate incomplete state information,and jointly opti-mized performance rewards and security costs,guiding the policy to improve network performance while proactively avoiding high-risk links and anomalous nodes.Simulation results show that GP-SR outperforms baseline algorithms in end-to-end delay,throughput,and packet delivery ratio,and demonstrates stronger robustness and adaptability under pro-gressively intensified attack scenarios.