arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2407.09251cs.LGcs.AIeess.SP
收藏

针对多级Lp攻击的深度对抗防御

Deep Adversarial Defense Against Multilevel-Lp Attacks

Illinois Institute of Technology(伊利诺伊理工学院);University of Michigan(密歇根大学)

Ren Wang, Yuxuan Li, Alfred Hero

AI总结:

针对传统对抗训练仅能防御单一范数攻击的局限,提出高效鲁棒模式连通性(EMRC)方法,融合ℓ₁与ℓ∞范数对抗训练解,实现多ℓ_p范数攻击防御,在多个数据集与模型架构上性能优于现有方法。

摘要:

深度学习模型已表现出对对抗攻击的显著脆弱性,尤其是在攻击者策略愈发复杂的情况下。传统对抗训练(AT)技术虽能提供一定的鲁棒性,但通常聚焦于防御单一类型的攻击,例如ℓ∞范数攻击,面对其他类型攻击时可能失效。本文提出了一种计算高效的多级ℓ_p防御方法,名为高效鲁棒模式连通性(Efficient Robust Mode Connectivity, EMRC)方法,旨在增强深度学习模型对多种ℓ_p范数攻击的抵御能力。该方法借鉴连续优化中的解析延拓思路,融合两个针对特定p值的对抗最优模型——即ℓ₁范数和ℓ∞范数的对抗训练解——为一定p值范围内的攻击提供良好的对抗鲁棒性。我们的实验表明,在CIFAR-10、CIFAR-100数据集以及PreResNet110、WideResNet、ViT-Base架构上,该方法相较于AT-ℓ∞、E-AT和MSD方法,在各类攻击下表现更优。

英文摘要:

Deep learning models have shown considerable vulnerability to adversarial attacks, particularly as attacker strategies become more sophisticated. While traditional adversarial training (AT) techniques offer some resilience, they often focus on defending against a single type of attack, e.g., the $\ell_\infty$-norm attack, which can fail for other types. This paper introduces a computationally efficient multilevel $\ell_p$ defense, called the Efficient Robust Mode Connectivity (EMRC) method, which aims to enhance a deep learning model's resilience against multiple $\ell_p$-norm attacks. Similar to analytical continuation approaches used in continuous optimization, the method blends two $p$-specific adversarially optimal models, the $\ell_1$- and $\ell_\infty$-norm AT solutions, to provide good adversarial robustness for a range of $p$. We present experiments demonstrating that our approach performs better on various attacks as compared to AT-$\ell_\infty$, E-AT, and MSD, for datasets/architectures including: CIFAR-10, CIFAR-100 / PreResNet110, WideResNet, ViT-Base.

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。

↑