返回全部资讯
OpenAI 官方博客T1·基座大模型

研究者展示多视角鲁棒对抗样本,挑战自动驾驶安全论断

原标题:Robust adversarial inputs
45AI 研判分

研究人员成功构建了能在不同尺度、角度和视角下稳定欺骗神经网络分类器的图像。这一成果直接反驳了上周关于“多视角采集使自动驾驶汽车难以被恶意攻击”的观点,揭示了现有防御机制的局限性。

报道正文

We’ve created images that reliably fool neural network classifiers when viewed from varied scales and perspectives. This challenges a claim from last week that self-driving cars would be hard to trick maliciously since they capture images from multiple scales, angles, perspectives, and the like.