arXivDaily arXiv每日学术速递 周一至周五更新
北美退伍军人节(Veterans Day),12 日无更新
arXiv 1909.04068cs.LGcs.AIstat.ML
收藏

多个扰动模型联合的对抗鲁棒性

Adversarial Robustness Against the Union of Multiple Perturbation Models

IIT Delhi(印度理工学院德里分校);Carnegie Mellon University(卡内基梅隆大学);Bosch Center for Artificial Intelligence(博世人工智能中心)

Pratyush Maini, Eric Wong, J. Zico Kolter

摘要:

由于深度学习系统对对抗攻击的敏感性,在开发(经验上和可验证的)鲁棒分类器方面已经做了大量工作。虽然大多数工作都是针对单一类型的攻击进行防御,但最近的工作着眼于使用多个攻击的简单聚合来防御多个扰动模型。然而,这些方法可能难以调整,并且可能容易导致对各个扰动模型的鲁棒性的不平衡程度,从而导致联合上的次优最坏情况损失。在这项工作中,我们开发了一个自然的推广标准的基于PGD的程序,将多个扰动模型到一个单一的攻击,采取最坏的情况下,在所有最陡的下降方向。这种方法的优点是直接收敛在不同的扰动模型之间的权衡,最大限度地减少了工会的最坏情况下的性能。通过这种方法,我们能够训练标准体系结构,这些体系结构同时对$\ell_\infty$、$\ell_2$和$\ell_1$攻击具有鲁棒性,在MNIST和CIFAR10数据集上的性能优于过去的方法,并实现了47.0%的对抗准确度。($\ell_\infty$,$\ell_2$,$\ell_1$)扰动,其半径=(0.03,0.5,12),改进了先前的方法,其实现了40.6%的准确度。

英文摘要:

Owing to the susceptibility of deep learning systems to adversarial attacks, there has been a great deal of work in developing (both empirically and certifiably) robust classifiers. While most work has defended against a single type of attack, recent work has looked at defending against multiple perturbation models using simple aggregations of multiple attacks. However, these methods can be difficult to tune, and can easily result in imbalanced degrees of robustness to individual perturbation models, resulting in a sub-optimal worst-case loss over the union. In this work, we develop a natural generalization of the standard PGD-based procedure to incorporate multiple perturbation models into a single attack, by taking the worst-case over all steepest descent directions. This approach has the advantage of directly converging upon a trade-off between different perturbation models which minimizes the worst-case performance over the union. With this approach, we are able to train standard architectures which are simultaneously robust against $\ell_\infty$, $\ell_2$, and $\ell_1$ attacks, outperforming past approaches on the MNIST and CIFAR10 datasets and achieving adversarial accuracy of 47.0% against the union of ($\ell_\infty$, $\ell_2$, $\ell_1$) perturbations with radius = (0.03, 0.5, 12) on the latter, improving upon previous approaches which achieve 40.6% accuracy.

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。

↑