文档图像分类的对抗鲁棒性评估
Evaluating Adversarial Robustness on Document Image Classification
AI总结：
针对现有对抗攻防研究多聚焦普通图像、较少涉及文档图像的问题，该研究将三类无目标对抗攻击应用于文档与自然数据，评估三种防御手段对两种模型在文档图像分类任务上的鲁棒性影响，填补了相关研究空白。
摘要：
近年来，对抗攻击与防御在计算机视觉系统领域受到越来越多的关注，但截至目前，多数研究仅局限于普通图像。然而，许多人工智能模型实际处理的是文档类数据，这类数据与真实世界图像存在显著差异。因此，本研究尝试将对抗攻击思路应用于文档数据与自然数据，并提升模型对此类攻击的防御能力。我们聚焦于无目标的基于梯度、基于迁移和基于分数的攻击方式，评估对抗训练、JPEG输入压缩及灰度输入转换对ResNet50和EfficientNetB0模型架构鲁棒性的影响。据我们所知，学界尚未开展此类研究来探讨这些攻击对文档图像分类任务的影响。
英文摘要：
Adversarial attacks and defenses have gained increasing interest on computer vision systems in recent years, but as of today, most investigations are limited to images. However, many artificial intelligence models actually handle documentary data, which is very different from real world images. Hence, in this work, we try to apply the adversarial attack philosophy on documentary and natural data and to protect models against such attacks. We focus our work on untargeted gradient-based, transfer-based and score-based attacks and evaluate the impact of adversarial training, JPEG input compression and grey-scale input transformation on the robustness of ResNet50 and EfficientNetB0 model architectures. To the best of our knowledge, no such work has been conducted by the community in order to study the impact of these attacks on the document image classification task.
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。
