arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2202.01181cs.LGcs.CV
收藏

Make Some Noise:可靠高效的单步的对抗训练

Make Some Noise: Reliable and Efficient Single-Step Adversarial Training

University of Oxford(牛津大学);NAVER LABS Europe(NAVER LABS欧洲);ETH Zürich(苏黎世联邦理工学院);ETH AI Center(苏黎世联邦理工学院人工智能中心);Five AI Ltd.(Five AI有限公司)

Pau de Jorge, Adel Bibi, Riccardo Volpi, Amartya Sanyal, Philip H. S. Torr, Grégory Rogez, Puneet K. Dokania

摘要:

最近,Wong等人表明,使用单步的【形】FGSM进行对抗性训练会导致一种名为灾难性过拟合(CO)的特征故障模式,其中模型突然变得容易受到多步攻击。然而,Andriushchenko和Flammarion观察到RS-FGSM在较大的扰动下仍然会导致CO,并提出了一种计算代价高昂的正则化器(GradAlign)来避免它。在这项工作中,我们系统地重新审视了噪声和裁剪在单步的【形】对抗训练中的作用。与以前的直觉相反,我们发现,使用更强的噪声周围的干净的样品结合\textit{not clipping}是非常有效的,以避免CO大扰动半径。然后,我们提出了Noise-FGSM(N-FGSM),在提供单步的【形】对抗训练的好处的同时,不会受到CO的影响。对大量实验的实证分析表明,N-FGSM能够匹配或超过以前最先进的GradAlign的性能,同时实现3倍的速度提升。代码可以在https://github.com/pdejorge/N-FGSM中找到

英文摘要:

Recently, Wong et al. showed that adversarial training with single-step FGSM leads to a characteristic failure mode named Catastrophic Overfitting (CO), in which a model becomes suddenly vulnerable to multi-step attacks. Experimentally they showed that simply adding a random perturbation prior to FGSM (RS-FGSM) could prevent CO. However, Andriushchenko and Flammarion observed that RS-FGSM still leads to CO for larger perturbations, and proposed a computationally expensive regularizer (GradAlign) to avoid it. In this work, we methodically revisit the role of noise and clipping in single-step adversarial training. Contrary to previous intuitions, we find that using a stronger noise around the clean sample combined with \textit{not clipping} is highly effective in avoiding CO for large perturbation radii. We then propose Noise-FGSM (N-FGSM) that, while providing the benefits of single-step adversarial training, does not suffer from CO. Empirical analyses on a large suite of experiments show that N-FGSM is able to match or surpass the performance of previous state-of-the-art GradAlign, while achieving 3x speed-up. Code can be found in https://github.com/pdejorge/N-FGSM

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。

↑