arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 1910.09239cs.LGstat.ML
收藏

恢复本地化对抗攻击

Recovering Localized Adversarial Attacks

Bielefeld University(比勒费尔德大学);Honda Research Institue Europe(本田欧洲研究院)

Jan Philip Göpfert, Heiko Wersing, Barbara Hammer

摘要:

近年来,深度卷积神经网络取得了巨大的成功,特别是在计算机视觉领域。它们快速、方便,而且--由于成熟的框架--相对容易实现和部署。然而,他们的推理隐藏在一个黑盒子里,尽管有许多建议的方法试图为神经网络的预测提供人类可以理解的解释。这仍然是一个争论的问题,这些解释者是最适合的情况下,以及如何定量评估和比较。在这篇文章中,我们关注的是在极端情况下卷积深度神经网络的解释器的能力:在这种情况下,人类和网络从根本上不同意。深度神经网络容易受到对抗攻击的影响,这些攻击会故意修改输入样本以误导神经网络的分类,而不会影响人类观察者如何解释输入。我们的目标是通过调查他们是否可以识别图像的adversarially攻击区域来评估解释者。特别是,我们定量和定性调查的能力,三个流行的分类解释-经典的显着性,引导反向传播,石灰-相对于他们的能力,以确定攻击区域的解释区域的(不正确的)预测在代表性的例子从图像分类。我们发现LIME优于其他解释器。

英文摘要:

Deep convolutional neural networks have achieved great successes over recent years, particularly in the domain of computer vision. They are fast, convenient, and -- thanks to mature frameworks -- relatively easy to implement and deploy. However, their reasoning is hidden inside a black box, in spite of a number of proposed approaches that try to provide human-understandable explanations for the predictions of neural networks. It is still a matter of debate which of these explainers are best suited for which situations, and how to quantitatively evaluate and compare them. In this contribution, we focus on the capabilities of explainers for convolutional deep neural networks in an extreme situation: a setting in which humans and networks fundamentally disagree. Deep neural networks are susceptible to adversarial attacks that deliberately modify input samples to mislead a neural network's classification, without affecting how a human observer interprets the input. Our goal with this contribution is to evaluate explainers by investigating whether they can identify adversarially attacked regions of an image. In particular, we quantitatively and qualitatively investigate the capability of three popular explainers of classifications -- classic salience, guided backpropagation, and LIME -- with respect to their ability to identify regions of attack as the explanatory regions for the (incorrect) prediction in representative examples from image classification. We find that LIME outperforms the other explainers.

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。

↑