Security and Privacy Issues in Deep Reinforcement Learning: Threats and Countermeasures

K Mo, P Ye, X Ren, S Wang, W Li, J Li - ACM Computing Surveys, 2024 - dl.acm.org
Deep Reinforcement Learning (DRL) is an essential subfield of Artificial Intelligence (AI),
where agents interact with environments to learn policies for solving complex tasks. In recent …

Certified policy smoothing for cooperative multi-agent reinforcement learning

R Mu, W Ruan, LS Marcolino, G Jin, Q Ni - Proceedings of the AAAI …, 2023 - ojs.aaai.org
Cooperative multi-agent reinforcement learning (c-MARL) is widely applied in safety-critical
scenarios, thus the analysis of robustness for c-MARL models is profoundly important …

Robustness testing for multi-agent reinforcement learning: State perturbations on critical agents

Z Zhou, G Liu - arXiv preprint arXiv:2306.06136, 2023 - arxiv.org
Multi-Agent Reinforcement Learning (MARL) has been widely applied in many fields such
as smart traffic and unmanned aerial vehicles. However, most MARL algorithms are …

Enhancing the robustness of qmix against state-adversarial attacks

W Guo, G Liu, Z Zhou, L Wang, J Wang - Neurocomputing, 2024 - Elsevier
Abstract Multi-Agent Reinforcement Learning (MARL) trains the decision models of
cooperative agents by making them gain the highest rewards. The Centralized Training with …

Multiagent Reinforcement Learning: Methods, Trustworthiness, Applications in Intelligent Vehicles, and Challenges

Z Zhou, G Liu, Y Tang - IEEE Transactions on Intelligent …, 2024 - ieeexplore.ieee.org
Multiagent Reinforcement Learning (MARL) plays a pivotal role in intelligent vehicle
systems, offering solutions for complex decision-making, coordination, and adaptive …

SleeperNets: Universal Backdoor Poisoning Attacks Against Reinforcement Learning Agents

E Rathbun, C Amato, A Oprea - arXiv preprint arXiv:2405.20539, 2024 - arxiv.org
Reinforcement learning (RL) is an actively growing field that is seeing increased usage in
real-world, safety-critical applications--making it paramount to ensure the robustness of RL …