arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 2301.12554cs.LGcs.CRcs.CV
收藏

通过自适应平滑改善分类器的精度-鲁棒性权衡

Improving the Accuracy-Robustness Trade-Off of Classifiers via Adaptive Smoothing

University of California, Berkeley(加利福尼亚大学伯克利分校);Scale AI

Yatong Bai, Brendon G. Anderson, Aerin Kim, Somayeh Sojoudi

AI总结:

本文提出自适应平滑方法,通过混合标准与鲁棒分类器并自适应调整权重,显著改善精度-鲁棒性权衡,在CIFAR-100上实现85.21%干净精度和38.72%鲁棒精度。

摘要:

尽管先前的研究提出了大量构建对对抗鲁棒性具有鲁棒性的神经分类器的方法,但从业者仍然不愿采用这些方法,因为它们会带来不可接受的严重干净精度损失。本文通过混合标准分类器和鲁棒分类器的输出概率,显著缓解了这种精度-鲁棒性权衡,其中标准网络针对干净精度进行优化,通常不具备鲁棒性。我们表明,鲁棒基分类器对正确和错误样本的置信度差异是这种改进的关键。除了提供直觉和实证证据外,我们还在现实假设下对混合分类器的鲁棒性进行了理论认证。此外,我们将对抗性输入检测器适配为混合网络,该网络自适应地调整两个基模型的混合比例,进一步减少实现鲁棒性所需的精度损失。所提出的灵活方法称为“自适应平滑”,可以与现有甚至未来的改善干净精度、鲁棒性或对抗检测的方法结合使用。我们的实证评估考虑了强攻击方法,包括AutoAttack和自适应攻击。在CIFAR-100数据集上,我们的方法实现了85.21%的干净精度,同时保持了38.72%的ℓ∞-AutoAttack(ε=8/255)精度,在提交时成为RobustBench CIFAR-100基准上第二鲁棒的方法,同时与所有列出的模型相比,干净精度提高了十个百分点。实现我们方法的代码可在https://github.com/Bai-YT/AdaptiveSmoothing获取。

英文摘要:

While prior research has proposed a plethora of methods that build neural classifiers robust against adversarial robustness, practitioners are still reluctant to adopt them due to their unacceptably severe clean accuracy penalties. This paper significantly alleviates this accuracy-robustness trade-off by mixing the output probabilities of a standard classifier and a robust classifier, where the standard network is optimized for clean accuracy and is not robust in general. We show that the robust base classifier's confidence difference for correct and incorrect examples is the key to this improvement. In addition to providing intuitions and empirical evidence, we theoretically certify the robustness of the mixed classifier under realistic assumptions. Furthermore, we adapt an adversarial input detector into a mixing network that adaptively adjusts the mixture of the two base models, further reducing the accuracy penalty of achieving robustness. The proposed flexible method, termed "adaptive smoothing", can work in conjunction with existing or even future methods that improve clean accuracy, robustness, or adversary detection. Our empirical evaluation considers strong attack methods, including AutoAttack and adaptive attack. On the CIFAR-100 dataset, our method achieves an 85.21% clean accuracy while maintaining a 38.72% $\ell_\infty$-AutoAttacked ($ε= 8/255$) accuracy, becoming the second most robust method on the RobustBench CIFAR-100 benchmark as of submission, while improving the clean accuracy by ten percentage points compared with all listed models. The code that implements our method is available at https://github.com/Bai-YT/AdaptiveSmoothing.

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。

↑