Security-Aware Virtual Network Embedding Algorithm Based on Reinforcement Learning

Virtual network embedding (VNE) algorithm is always the key problem in virtualization (NV) technology. At present, research this field still has following problems. The traditional way to solve VNE use heuristic algorithm. However, method relies on manual rules, which does not accord with actual situation of VNE. In addition, as intelligent learning become a trend, gradually outdated. same time, there are some security problems no For reason, paper proposes security-aware based reinforcement (RL). training phase, we policy agent and take extracted attributes substrate nodes form feature matrix input. trained environment get mapping probability each node. test map according breadth-first strategy (BFS) links. problem, add requirements level constraint for virtual node can only be embedded that lower than requirements. Experimental results show proposed superior other typical algorithms terms long-term average return, revenue consumption ratio request (VNR) acceptance rate.