Almost Tight L0-norm Certified Robustness of Top-k Predictions against Adversarial Perturbations

Almost Tight L0-norm Certified Robustness of Top-k Predictions against Adversarial Perturbations
复制标题

DOI:
--
复制
发表时间:
2020-11
期刊:
ArXiv
影响因子:
--
通讯作者:
Jinyuan Jia;Binghui Wang;Xiaoyu Cao;Hongbin Liu;N. Gong
Jinyuan Jia;Binghui Wang;Xiaoyu Cao;Hongbin Liu;N. Gong
中科院分区:
其他
文献类型:
--
作者:
Jinyuan Jia;Binghui Wang;Xiaoyu Cao;Hongbin Liu;N. Gong

文献摘要

相似文献

Top-$k$预测被用于许多现实世界的应用,如机器学习服务、推荐系统和网络搜索。$\ell_0 $-norm对抗扰动描述了任意修改输入的某些特征的攻击,使得分类器对扰动的输入做出不正确的预测。$\ell_0 $-norm对抗扰动很容易解释,并且可以在物理世界中实现。因此,证明top-$k$预测对$\ell_0$-norm对抗性扰动的鲁棒性是重要的。然而,现有的研究要么集中于证明前1 $预测的$\ell_0 $-范数稳健性,要么集中于前k$预测的$\ell_2 $-范数稳健性。在这项工作中,我们的目标是弥合差距。我们的方法是基于随机平滑,建立一个可证明的强大的分类器从任意分类器通过随机化的输入。我们的主要理论贡献是一个几乎紧密的$\ell_0 $规范认证的鲁棒性保证前$k$预测。我们在CIFAR 10和ImageNet上对我们的方法进行了实证评估。例如,我们的方法可以构建一个分类器,当攻击者可以任意干扰测试图像的5个像素时,该分类器在ImageNet上达到了69.2%的认证前3名准确率。
Top-$k$ predictions are used in many real-world applications such as machine learning as a service, recommender systems, and web searches. $\ell_0$-norm adversarial perturbation characterizes an attack that arbitrarily modifies some features of an input such that a classifier makes an incorrect prediction for the perturbed input. $\ell_0$-norm adversarial perturbation is easy to interpret and can be implemented in the physical world. Therefore, certifying robustness of top-$k$ predictions against $\ell_0$-norm adversarial perturbation is important. However, existing studies either focused on certifying $\ell_0$-norm robustness of top-$1$ predictions or $\ell_2$-norm robustness of top-$k$ predictions. In this work, we aim to bridge the gap. Our approach is based on randomized smoothing, which builds a provably robust classifier from an arbitrary classifier via randomizing an input. Our major theoretical contribution is an almost tight $\ell_0$-norm certified robustness guarantee for top-$k$ predictions. We empirically evaluate our method on CIFAR10 and ImageNet. For instance, our method can build a classifier that achieves a certified top-3 accuracy of 69.2\% on ImageNet when an attacker can arbitrarily perturb 5 pixels of a testing image.