scholar.google.com › citations
Sep 22, 2023 · In this work, we propose the selective real-time attack method in the deep reinforcement learning security domain, a new selective attack ...
Mar 16, 2024 · Recent work has shown that deep reinforcement learning (DRL) is vulnerable to adversarial attacks, so that exploiting vulnerabilities in DRL ...
Sep 22, 2023 · Comparative experiments show that our attack method can perform real‐time attacks while maintaining the attack effect and stealthiness.
Abstract. Recent work has shown that deep reinforcement learning (DRL) is vulnerable to adversarial attacks, so that exploiting vulnerabilities in DRL ...
Jun 16, 2021 · Deep reinforcement learning (DRL) is vulnerable to adversarial perturbations. Adversaries can mislead the policies of DRL agents by perturbing ...
Missing: Selective | Show results with:Selective
Sep 26, 2022 · Adversaries can mislead the policies of DRL agents by perturbing the state of the environment observed by the agents. Existing attacks are ...
Missing: Selective | Show results with:Selective
Adversaries can mislead the policies of DRL agents by perturbing the state of the environment observed by the agents. Existing attacks are feasible in principle ...
People also ask
What are adversarial perturbations?
What does reinforcement learning teach agents using?
Since the adversarial attacks at different time steps are not equally effective, this attack reduces the accumulated reward with fewer adversarial perturbations ...
We frame an adversarial attack as a streaming algorithm in which the victim learns a behavior based on the stream of its previous samples, while the adversary ...
Mar 20, 2023 · Robust deep reinforcement learning against adversarial perturbations on state observations. In. H. Larochelle, M. Ranzato, R. Hadsell, M. F. ...