LaserMark
收藏资源简介:
LaserMark是由上海科技大学和鹏城实验室共同创建的,用于支持后门攻击与防御研究的第一个真实世界交通标志数据集。该数据集包含带有物理激光点的交通标志图像,旨在模拟物理后门攻击场景。数据集的创建过程涉及在真实交通标志上使用激光点进行标记,并通过传感器捕捉这些标记的图像。LaserMark主要应用于深度神经网络的安全性研究,特别是针对自动驾驶中的交通标志识别系统,旨在解决物理后门攻击的检测与防御问题。
LaserMark is the first real-world traffic sign dataset for backdoor attack and defense research, co-developed by ShanghaiTech University and Peng Cheng Laboratory. It contains images of traffic signs marked with physical laser dots, designed to simulate physical backdoor attack scenarios. The dataset construction process involves marking real traffic signs with laser dots and capturing images of these marked signs using sensors. LaserMark is mainly used for safety research on deep neural networks, particularly traffic sign recognition systems in autonomous driving, with the objective of solving the detection and defense problems of physical backdoor attacks.

- 1LaserGuider: A Laser Based Physical Backdoor Attack against Deep Neural Networks上海科技大学 · 2024年



