QLSFC: An Intelligent Security Function Chain with Q-Learning in SDN/NFV Network.

5G suffers from new security and privacy issues. More specifically, distributed denial-of-service (DDoS) is one of the most common and destructive attack types. The traditional network does not perform well, because they do not quickly support the deployment of new service function as well as provide network security service on demand. Recently, with the rise of Reinforcement Learning (RL), Software Defined Network (SDN)/Network Functions Virtualization (NFV), and Interface to Network Security Functions (I2NSF) techniques, they play a significant role in security enhancement. Hence, we propose a framework to integrate RL with SDN/NFV and I2NSF architecture (RL-SDNV), in which the RL agent can intelligently select suitable security function chain (SFC) under attack scenarios. Then, we model the DDoS detection problem as a Markov Decision Process (MDP), and a QLearning detection algorithm (QLSFC) is proposed, in which the designed reward includes the processing time and the malicious traffic reduction rate. Finally, we build a corresponding prototype system and verify the feasibility and effectiveness of the proposed algorithm through compared experiments with other RL algorithms.