Application of deep reinforcement learning to intrusion detection for supervised problems

作者：

Highlights：

• Novel application of Reinforcement Learning (RL) to network intrusion detection.

• Novel application of RL to a labeled dataset.

• Detailed comparison of results with the most common machine learning models.

• The resulting detector is based on an extremely simple and fast policy network.

• The study includes 4 variants of RL algorithms and 2 intrusion detection datasets.

摘要

•Novel application of Reinforcement Learning (RL) to network intrusion detection.•Novel application of RL to a labeled dataset.•Detailed comparison of results with the most common machine learning models.•The resulting detector is based on an extremely simple and fast policy network.•The study includes 4 variants of RL algorithms and 2 intrusion detection datasets.

论文关键词：Intrusion detection,Data networks,Deep reinforcement learning

论文评审过程：Received 9 May 2019, Revised 26 July 2019, Accepted 17 September 2019, Available online 18 September 2019, Version of Record 26 September 2019.

论文官网地址：https://doi.org/10.1016/j.eswa.2019.112963