Minimalistic attacks : how little it takes to fool deep reinforcement learning policies

• Main Authors: Qu, Xinghua, Sun, Zhu, Ong, Yew-Soon, Gupta, Abhishek, Wei, Pengfei
• 其他作者: School of Computer Science and Engineering
• 格式: Article
• 語言: English
• 出版: 2021
• 主題: Engineering::Computer science and engineering; Reinforcement Learning; Adversarial Attacks
• 在線閱讀: https://hdl.handle.net/10356/153700
• 機構: Nanyang Technological University

Recent studies have revealed that neural network-based policies can be easily fooled by adversarial examples. However, while most prior works analyze the effects of perturbing every pixel of every frame assuming white-box policy access, in this paper we take a more restrictive view towards adversary...