Stop-and-Go：探索对基于深度强化学习的交通拥塞控制系统的后门攻击
Stop-and-Go: Exploring Backdoor Attacks on Deep Reinforcement Learning-based Traffic Congestion Control Systems
摘要：
最近的研究表明，在交通中引入自动驾驶汽车（AV）可以帮助减少交通堵塞。深度强化学习方法在复杂的控制问题中表现出良好的性能，包括自主车辆控制，并已用于最先进的AV控制器。然而，深度神经网络（DNN）使自动驾驶容易受到基于机器学习的攻击。在这项工作中，我们探讨了后门/木马的DRL为基础的AV控制器。我们开发了一个触发器的设计方法，是基于完善的交通物理学原理。恶意行为包括车辆减速和加速以导致出现走走停停的交通波（拥堵攻击）或AV加速导致AV撞到前面的车辆（保险攻击）。我们在单通道和双通道电路上测试我们的攻击。我们的实验结果表明，后门模型不损害正常的操作性能，与累积奖励的最大下降为1%。尽管如此，当相应的触发器出现时，它仍然可以被恶意激活以导致崩溃或拥塞。
英文摘要：
Recent work has shown that the introduction of autonomous vehicles (AVs) in traffic could help reduce traffic jams. Deep reinforcement learning methods demonstrate good performance in complex control problems, including autonomous vehicle control, and have been used in state-of-the-art AV controllers. However, deep neural networks (DNNs) render automated driving vulnerable to machine learning-based attacks. In this work, we explore the backdooring/trojanning of DRL-based AV controllers. We develop a trigger design methodology that is based on well-established principles of traffic physics. The malicious actions include vehicle deceleration and acceleration to cause stop-and-go traffic waves to emerge (congestion attacks) or AV acceleration resulting in the AV crashing into the vehicle in front (insurance attack). We test our attack on single-lane and two-lane circuits. Our experimental results show that the backdoored model does not compromise normal operation performance, with the maximum decrease in cumulative rewards being 1%. Still, it can be maliciously activated to cause a crash or congestion when the corresponding triggers appear.
机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。
