JSAI2024

Presentation information

General Session

General Session » GS-11 AI and Society

[2G1-GS-11] AI and Society:

Wed. May 29, 2024 9:00 AM - 10:40 AM Room G (Room 22+23)

座長:髙橋 翼(LINEヤフー/SB Intuitions)

9:40 AM - 10:00 AM

[2G1-GS-11-03] Targeted Manipulation Attacks on Reinforcement Learning Agents through Imitation Learning with Perturbed Observations

〇Shojiro Yamabe1, Kazuto Fukuchi2,3, Ryoma Senda3, Jun Sakuma1,3 (1. Tokyo Institute of Technology, 2. University of Tsukuba, 3. RIKEN AIP)

Keywords:Deep Reinforcement Learning, Adversarial attack, Generative adversarial networks

Deep reinforcement learning (DRL) is known to be vulnerable to adversarial attacks. For real-world applications, it is necessary to improve the robustness of DRL agents. Therefore, in this study, we propose a targeted manipulation attack method that specifies the behavior of the victim agent assuming a real-world attack in order to investigate the vulnerability. As a threat model, we consider a situation in which an attacker can generate perturbations to the observations of victim agent. The goal of the attacker is to manipulate the victim agent. The attacker expresses the desired behavior as a trajectory and attacks the victim agent to imitate it. In this study, we use imitation learning to realize the attack. Finally, we confirm that the targeted manipulation attack succeeds under the threat model set by our experiments on MetaWorld, a benchmark for reinforcement learning.

Please log in with your participant account.
» Participant Log In