Deep-attack over the deep reinforcement learning

作者：

Highlights：

• Adversarial attack over DRL in terms of attack efficiency and stealth is explored.

• A set of evaluation metrics with regard to attack efficiency and stealth is proposed.

• The validity of the attack method and measurements is verified by experiments.

摘要

•Adversarial attack over DRL in terms of attack efficiency and stealth is explored.•A set of evaluation metrics with regard to attack efficiency and stealth is proposed.•The validity of the attack method and measurements is verified by experiments.

论文关键词：Adversarial attack,Deep reinforcement learning,Adversarial training

论文评审过程：Received 22 July 2021, Revised 29 April 2022, Accepted 30 April 2022, Available online 10 May 2022, Version of Record 27 May 2022.

论文官网地址：https://doi.org/10.1016/j.knosys.2022.108965