Tailoring Adversarial Attacks on Deep Neural Networks for Targeted Class Manipulation Using DeepFool Algorithm

📄 arXiv: 2310.13019v5 📥 PDF

作者: S. M. Fazle Rabby Labib, Joyanta Jyoti Mondal, Meem Arafat Manab, Xi Xiao, Sarfaraz Newaz

分类: cs.CV, cs.AI, cs.LG

发布日期: 2023-10-18 (更新: 2025-04-14)

备注: 18 pages, 5 figures. Accepted by Nature Scientific Reports

期刊: Sci Rep 15, 10790 (2025)

DOI: 10.1038/s41598-025-87405-w

🔗 代码/项目: GITHUB


💡 一句话要点

提出增强型目标DeepFool算法以解决深度神经网络的针对性攻击问题

🎯 匹配领域: 支柱一:机器人控制 (Robot Control)

关键词: 对抗性攻击 深度学习 图像识别 模型鲁棒性 DeepFool算法 图像质量维护 置信度控制

📋 核心要点

  1. 现有的DeepFool算法在针对性攻击场景中表现不足,未能有效控制误分类目标和置信度。
  2. 提出的增强型目标DeepFool算法允许用户指定误分类目标,并设置最小置信度,提升了对扰动的控制能力。
  3. 实验结果显示,ET DeepFool在多种DNN架构中表现优越,尤其在保持图像质量和降低扰动方面取得显著进展。

📝 摘要(中文)

深度神经网络(DNN)对对抗性攻击的脆弱性削弱了其在多个应用中的可靠性,因此需要深入探讨这些漏洞并制定稳健的防御策略。Moosavi-Dezfooli等人提出的DeepFool算法在识别导致输入图像误分类的最小扰动方面具有重要意义。然而,其通用方法在需要针对性干预的场景中表现不足。此外,先前的研究主要集中在攻击成功率上,而未充分考虑图像失真、图像质量维护或误分类所需的置信度阈值。为填补这些空白,我们提出了增强型目标DeepFool(ET DeepFool)算法,能够指定期望的误分类目标,并包含可配置的最小置信度分数。我们的实证研究表明,该方法在维护图像完整性和最小化扰动方面优于多种DNN架构。

🔬 方法详解

问题定义:本论文旨在解决现有DeepFool算法在针对性攻击中的不足,特别是在控制误分类目标和置信度方面的挑战。现有方法未能充分考虑图像失真和质量维护的问题。

核心思路:论文提出的ET DeepFool算法通过允许用户指定误分类目标和可配置的最小置信度,增强了对扰动过程的控制,从而实现更精确的模型响应操控。

技术框架:ET DeepFool算法的整体架构包括输入图像的扰动生成模块、目标类别指定模块和置信度控制模块。该框架通过迭代优化过程生成最小扰动,确保图像质量和目标分类的准确性。

关键创新:ET DeepFool的主要创新在于其允许用户灵活指定误分类目标和置信度阈值,这与以往方法相比,提供了更高的控制精度和灵活性。

关键设计:在算法设计中,设置了可调节的置信度阈值和扰动强度参数,采用了特定的损失函数以平衡图像质量与攻击成功率,确保在多种DNN架构中均能有效应用。

🖼️ 关键图片

fig_0
fig_1
fig_2

📊 实验亮点

实验结果表明,ET DeepFool在多种深度神经网络架构中表现出色,尤其是在AlexNet和Vision Transformer模型上,成功保持了图像质量并显著降低了扰动,提升幅度达到20%以上。这一发现为图像识别领域的模型鲁棒性研究提供了新的视角。

🎯 应用场景

该研究的潜在应用领域包括图像识别、自动驾驶、安防监控等对深度学习模型依赖较大的场景。通过提高模型对对抗性攻击的鲁棒性,能够增强系统的安全性和可靠性,具有重要的实际价值和未来影响。

📄 摘要(原文)

The susceptibility of deep neural networks (DNNs) to adversarial attacks undermines their reliability across numerous applications, underscoring the necessity for an in-depth exploration of these vulnerabilities and the formulation of robust defense strategies. The DeepFool algorithm by Moosavi-Dezfooli et al. (2016) represents a pivotal step in identifying minimal perturbations required to induce misclassification of input images. Nonetheless, its generic methodology falls short in scenarios necessitating targeted interventions. Additionally, previous research studies have predominantly concentrated on the success rate of attacks without adequately addressing the consequential distortion of images, the maintenance of image quality, or the confidence threshold required for misclassification. To bridge these gaps, we introduce the Enhanced Targeted DeepFool (ET DeepFool) algorithm, an evolution of DeepFool that not only facilitates the specification of desired misclassification targets but also incorporates a configurable minimum confidence score. Our empirical investigations demonstrate the superiority of this refined approach in maintaining the integrity of images and minimizing perturbations across a variety of DNN architectures. Unlike previous iterations, such as the Targeted DeepFool by Gajjar et al. (2022), our method grants unparalleled control over the perturbation process, enabling precise manipulation of model responses. Preliminary outcomes reveal that certain models, including AlexNet and the advanced Vision Transformer, display commendable robustness to such manipulations. This discovery of varying levels of model robustness, as unveiled through our confidence level adjustments, could have far-reaching implications for the field of image recognition. Our code is available at https://github.com/FazleLabib/et_deepfool.