Exploring and Exploiting Decision Boundary Dynamics for Adversarial Robustness
Exploring and Exploiting Decision Boundary Dynamics for Adversarial Robustness
复制标题
DOI：
10.48550/arxiv.2302.03015
复制
发表时间：
2023-02
期刊：
影响因子：
--
通讯作者：
Yuancheng Xu;Yanchao Sun;Micah Goldblum;T. Goldstein;Furong Huang
中科院分区：
文献类型：
--
作者：
Yuancheng Xu;Yanchao Sun;Micah Goldblum;T. Goldstein;Furong Huang
深度分类器的鲁棒性可以通过它的边缘来表征：决策边界到自然数据点的距离。然而，目前尚不清楚现有的稳健训练方法是否有效地增加了训练过程中每个脆弱点的裕度。为了理解这一点，我们提出了一个连续时间框架来量化决策边界相对于每个单独点的相对速度。通过可视化最有效的鲁棒训练算法之一——对抗训练中决策边界的移动速度，揭示了一种令人惊讶的移动行为：决策边界远离一些脆弱点，但同时又靠近另一些脆弱点，从而减少了它们的边缘。为了缓解决策边界的这些冲突动态，我们提出了动态感知鲁棒训练（DyART），它鼓励决策边界参与优先增加较小边际的运动。与之前的工作相比，DyART直接在边缘操作，而不是间接近似，从而允许更有针对性和有效的鲁棒性改进。在CIFAR-10和Tiny-ImageNet数据集上的实验验证了DyART缓解了决策边界的冲突动态，并且与最先进的防御相比，在各种扰动大小下获得了更好的鲁棒性。我们的代码可在https://github.com/Yuancheng-Xu/Dynamics-Aware-Robust-Training上获得。
The robustness of a deep classifier can be characterized by its margins: the decision boundary's distances to natural data points. However, it is unclear whether existing robust training methods effectively increase the margin for each vulnerable point during training. To understand this, we propose a continuous-time framework for quantifying the relative speed of the decision boundary with respect to each individual point. Through visualizing the moving speed of the decision boundary under Adversarial Training, one of the most effective robust training algorithms, a surprising moving-behavior is revealed: the decision boundary moves away from some vulnerable points but simultaneously moves closer to others, decreasing their margins. To alleviate these conflicting dynamics of the decision boundary, we propose Dynamics-aware Robust Training (DyART), which encourages the decision boundary to engage in movement that prioritizes increasing smaller margins. In contrast to prior works, DyART directly operates on the margins rather than their indirect approximations, allowing for more targeted and effective robustness improvement. Experiments on the CIFAR-10 and Tiny-ImageNet datasets verify that DyART alleviates the conflicting dynamics of the decision boundary and obtains improved robustness under various perturbation sizes compared to the state-of-the-art defenses. Our code is available at https://github.com/Yuancheng-Xu/Dynamics-Aware-Robust-Training.