衰减步长的力量：提升基于符号优化器的攻击稳定性与迁移性
The Power of Decaying Steps: Enhancing Attack Stability and Transferability for Sign-based Optimizers
把基于符号的对抗攻击优化器重新表述为逐坐标梯度下降，并用单调递减的坐标级步长（MDCS）解决迭代次数增加反而导致攻击成功率下降的问题；做迁移攻击或对抗训练的人可以直接把这一调度替换进 I-FGSM/MI-FGSM 类流程，提升黑盒迁移性与稳定性。
原文依据 we propose a series of new attack algorithms that enforce Monotonically Decreasing Coordinate-wise Step-sizes (MDCS) within sign-based optimizers
摘要译文
由模型翻译 · 术语统一 · 请对照原文核对关键结论对抗样本的构造可以被形式化为一个优化问题。尽管 I-FGSM 和 MI-FGSM 等基于符号的优化器已成为此类诱导优化问题事实上的标准，但在理论依据与实际可靠性方面仍存在若干未解决的问题，尤其是不收敛与不稳定，这不可避免地影响了其迁移性。与预期相反，我们观察到当执行更多迭代次数时，攻击成功率可能急剧下降。本文从优化视角处理这些问题。通过将基于符号的优化器重新表述为一种特定的逐坐标梯度下降，我们认为不收敛与不稳定的一个原因在于其非衰减的步长调度。基于这一观点，我们提出了一系列新的攻击算法，在基于符号的优化器内强制采用单调递减的坐标级步长（MDCS）。典型地，我们进一步给出理论保证，证明 MDCS-MI 达到 O(1/\sqrt T ) 的最优收敛速率，其中 T 为迭代次数。在图像分类与跨模态检索任务上的大量实验表明，与最优的基于符号的方法相比，我们的方法不仅显著提升了迁移性，还增强了攻击稳定性。
英文原文摘要 1377 字符
Crafting adversarial examples can be formulated as an optimization problem. While sign-based optimizers such as I-FGSM and MI-FGSM have become the de facto standard for the induced optimization problems, there still exist several unsolved problems in theoretical grounding and practical reliability especially in non-convergence and instability, which inevitably influences their transferability. Contrary to the expectation, we observe that the attack success rate may degrade sharply when more number of iterations are conducted. In this paper, we address these issues from an optimization perspective. By reformulating the sign-based optimizer as a specific coordinate-wise gradient descent, we argue that one cause for non-convergence and instability is their non-decaying step-size scheduling. Based upon this viewpoint, we propose a series of new attack algorithms that enforce Monotonically Decreasing Coordinate-wise Step-sizes (MDCS) within sign-based optimizers. Typically, we further provide theoretical guarantees proving that MDCS-MI attains an optimal convergence rate of O(1/\sqrt T ), where T is the number of iterations. Extensive experiments on image classification and cross-modal retrieval tasks demonstrate that our approach not only significantly improves transferability but also enhances attack stability compared to state-of-the-art sign-based methods.