Research map / Adversarial robustness and security / Black-box and transfer attacks
Frequency Domain Model Augmentation for Adversarial Attack
Yuyang Long, Qilong Zhang, Boheng Zeng, Lianli Gao, Xianglong Liu, Jian Zhang, Jingkuan Song
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add spectrum transformation (frequency-domain augmentation) to MI-FGSM | MI-FGSM · Normal models, crafted on Inc-v3 | avg success rate: 54.5 -> 63.9 (S2I-FGSM); 81.7 (FI-MI) -> 86.6 | Improved |
| add spectrum transformation to TI-DIM | TI-DIM · Nine defenses, crafted on Inc-v3 | avg success rate: 39.0 -> 74.7 | Improved |
| add spectrum transformation to SI-TI-DIM | S2I-SI-TI-DIM · Nine defenses, ensemble | avg success rate: 89.8 -> 95.4 | Improved |
| apply transformation in frequency domain rather than spatial domain (ablation) | S2I-FGSM · Normal models, Inc-v3 | success rate: +15.0% | Improved |
| add Gaussian noise xi and tuning matrix M (ablation) | spectrum transformation · Six normal models | avg success rate | Improved |
| apply DCT on whole image instead of small blocks (ablation) | S2I-FGSM · Inc-v3 | success rate | Improved |
| increase number of spectrum transformations N from 1 to 20 (ablation) | S2I-FGSM · Inc-v3 | success rate | Improved |
| increase tuning factor rho beyond 0.5 (ablation) | S2I-FGSM · Inc-v3 | success rate | Hurt |
| increase noise std sigma beyond 16 (ablation) | S2I-FGSM · Inc-v3 | success rate | Hurt |
| combine spectrum transformation with FI/VT-style MI-FGSM | FI-MI-FGSM · Six normal models | avg success rate: 81.7 -> 86.6 | Improved |
Closest papers by meaning
- FACL-Attack: Frequency-Aware Contrastive Learning for Transferable Adversarial Attacks (AAAI 2024 · similarity 0.88)
- Towards Transferable Adversarial Attacks with Centralized Perturbation (AAAI 2024 · similarity 0.87)
- Beyond ImageNet Attack: Towards Crafting Adversarial Examples for Black-box Domains (ICLR 2022 · similarity 0.87)
- Boosting Adversarial Transferability through Augmentation in Hypothesis Space (CVPR 2025 · similarity 0.86)
- Seeking Flat Minima Over Diverse Surrogates for Improved Adversarial Transferability: A Theoretical Framework and Algorithmic Instantiation (TPAMI 2026 · similarity 0.86)
- PubDef: Defending Against Transfer Attacks From Public Models (ICLR 2024 · similarity 0.85)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
