BriefGPT.xyz
Mar, 2020
动态刺激采样器: 基于动态刺激先验的视频对抗攻击
Motion-Excited Sampler: Video Adversarial Attack with Sparked Prior
HTML
PDF
Hu Zhang, Linchao Zhu, Yi Zhu, Yi Yang
TL;DR
本文旨在通过利用视频帧之间的内在运动模式和区域相对运动，攻击视频模型。我们提出了一种有效的运动激发采样器来获取运动感知噪声先验，我们将其称为激发先验。我们的激发先验强调帧之间的相关性并利用相对运动。通过在梯度估计中使用激发先验，我们成功地攻击了各种视频分类模型。
Abstract
Deep neural networks are known to be susceptible to
adversarial noise
, which are tiny and imperceptible perturbations. Most of previous work on adversarial attack mainly focus on image models, while the vulnerability of
→