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 2403.04070cs.LGcs.AIcs.CRcs.CV
收藏

使用感知脆弱性的扰动预算提升对抗训练效果

Improving Adversarial Training using Vulnerability-Aware Perturbation Budget

Iowa State University(爱荷华州立大学)

Olukorede Fakorede, Modeste Atsague, Jin Tian

AI总结:

针对固定扰动预算限制对抗训练效能的问题,提出两种基于样本脆弱性分配扰动半径的轻量重加权方法,有效提升了模型对多种对抗攻击的鲁棒性。

摘要:

对抗训练(Adversarial Training, AT)可有效提升深度神经网络(Deep Neural Networks, DNNs)对对抗攻击的鲁棒性。通常,AT 利用在预定义的固定扰动范围内生成的对抗样本训练 DNN 模型。值得注意的是,用于生成这些对抗样本的各个自然样本本身具有不同程度的内在脆弱性,因此对所有样本使用固定扰动半径生成对抗样本,可能无法充分发挥 AT 的效能。受这一观察启发,我们提出两种简单、计算成本低的感知脆弱性重加权函数,用于为 AT 所用的对抗样本分配扰动边界,分别名为 Margin-Weighted Perturbation Budget(MWPB,边距加权扰动预算)和 Standard-Deviation-Weighted Perturbation Budget(SDWPB,标准差加权扰动预算)。所提方法根据对应自然样本的脆弱性,为单个对抗样本分配扰动半径。实验结果表明,所提方法切实提升了 AT 算法抵御多种对抗攻击的鲁棒性。

英文摘要:

Adversarial Training (AT) effectively improves the robustness of Deep Neural Networks (DNNs) to adversarial attacks. Generally, AT involves training DNN models with adversarial examples obtained within a pre-defined, fixed perturbation bound. Notably, individual natural examples from which these adversarial examples are crafted exhibit varying degrees of intrinsic vulnerabilities, and as such, crafting adversarial examples with fixed perturbation radius for all instances may not sufficiently unleash the potency of AT. Motivated by this observation, we propose two simple, computationally cheap vulnerability-aware reweighting functions for assigning perturbation bounds to adversarial examples used for AT, named Margin-Weighted Perturbation Budget (MWPB) and Standard-Deviation-Weighted Perturbation Budget (SDWPB). The proposed methods assign perturbation radii to individual adversarial samples based on the vulnerability of their corresponding natural examples. Experimental results show that the proposed methods yield genuine improvements in the robustness of AT algorithms against various adversarial attacks.

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。

↑