信息安全研究 ›› 2022, Vol. 8 ›› Issue (3): 212-.

• 深度学习安全与对抗专题 • 上一篇    下一篇

针对深度强化学习导航的物理对抗攻击方法

桓琦;谢小权;郭敏;曾颖明;
  

  1. (中国航天科工集团第二研究院706所 北京100854

  • 出版日期:2022-03-01 发布日期:2022-03-01
  • 通讯作者: 桓琦 硕士研究生,主要研究方向为强化学习、人工智能安全 1209924748@qq.com
  • 作者简介:桓琦 硕士研究生,主要研究方向为强化学习、人工智能安全 1209924748@qq.com 谢小权 硕士、研究员,主要研究方向为信息安全 xiexiaoquan@163.com 郭敏 硕士、工程师,研究方向为人工智能安全 guominjmh@163.com 曾颖明 硕士、研究员,研究方向为网络安全 zengyingming@163.com

Physical Adversarial Attacks Against Deep Reinforcement Learning Based Navigation

  • Online:2022-03-01 Published:2022-03-01

摘要: 本文针对基于深度强化学习(deep reinforcement learning, DRL)的激光导航系统的安全性进行研究,首次提出了对抗地图的概念,并在此基础上提出了一种物理对抗攻击方法.该方法使用对抗样本生成算法计算激光测距传感器上的对抗扰动,然后修改原始地图实现这些扰动,得到对抗地图.对抗地图可以在某个特定区域诱导智能体偏离最优路径,最终使机器人导航失败.在物理仿真实验中,本文对比了智能体在多个原始地图和对抗地图的导航结果,证明了对抗地图攻击方法的有效性,也指出了目前DRL技术应用在导航系统上存在的安全隐患.

关键词: 深度强化学习, 自主导航, 对抗攻击, 对抗样本, 深度学习

Abstract: In this paper, the security of deep reinforcement learning (DRL) based laser navigation system is studied, and the concept of adversarial map and a physical attack method based on it is proposed for the first time. The method uses the adversarial example generation algorithm to calculate the noise on the laser sensor and then modifies the original map to realize these noises and get the adversarial map. The adversarial map can induce the agent to deviate from the optimal path in a particular area and finally makes the robot navigation fail. In the physical simulation experiment, this paper compares the navigation results of an agent in multiple original maps and adversarial maps, proves the effectiveness of the countermeasure map attack method, and points out the hidden security dangers of the current application of DRL technology in the navigation system.

Key words: deep reinforcement learning, autonomous navigation, artificial intelligence, adversarial attack, deep learning