Research map / Adversarial robustness and security / Black-box and transfer attacks
Improving Adversarial Transferability via Intermediate-level Perturbation Decay
Qizhang Li, Yiwen Guo, Wangmeng Zuo, Hao Chen
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 ILPD to I-FGSM | I-FGSM · ImageNet | average attack success rate: 22.36 -> 57.06 | Improved |
| add ILPD to I-FGSM | I-FGSM · CIFAR-10 | average attack success rate: 48.62 -> 62.82 | Improved |
| use ILPD instead of second-best ILA++ | I-FGSM · ImageNet | average attack success rate: +10.07 | Improved |
| add ILPD to MI-FGSM | MI-FGSM · ImageNet | average attack success rate: 28.57 -> 59.84 | Improved |
| add ILPD to DI$^2$-FGSM | DI$^2$-FGSM · ImageNet | average attack success rate: 46.88 -> 63.23 | Improved |
| add ILPD to Admix | Admix · ImageNet | average attack success rate: 35.59 -> 60.28 | Improved |
| add ILPD, attack robust models | I-FGSM · ImageNet robust victims | average attack success rate: 10.56 -> 24.66 | Improved |
| add ILPD vs NAA, attack smoothed ResNet-50 | I-FGSM · random smoothing ResNet-50 | attack success rate: 7.20 -> 12.90 | Improved |
| add ILPD, attack NRP-purified images | I-FGSM · NRP defense | average attack success rate: 9.39 -> 20.03 | Improved |
| use ILPD with ViT-B substitute instead of NAA | ILA++/NAA · ImageNet | average attack success rate: 43.52 -> 46.63 | Improved |
| use ILPD with Swin-B substitute instead of NAA | ILA++/NAA · ImageNet | average attack success rate: 21.41 -> 36.77 | Improved |
| choose the split position at the first layer (ablation) | ILPD · ImageNet | average attack success rate: 53.12 | No clear effect |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- Admix — Admix: Enhancing the Transferability of Adversarial Attacks (ICCV 2021) · better on 3 of 3 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Enhancing Adversarial Example Transferability With an Intermediate Level Attack (ICCV 2019 · similarity 0.92)
- An Intermediate-level Attack Framework on The Basis of Linear Regression (TPAMI 2022 · similarity 0.91)
- Improving Adversarial Transferability with Local Perturbation Augmentation (CVPR 2026 · similarity 0.90)
- Towards Transferable Adversarial Attacks with Centralized Perturbation (AAAI 2024 · similarity 0.89)
- Boosting Adversarial Transferability through Augmentation in Hypothesis Space (CVPR 2025 · similarity 0.88)
- Enhancing Adversarial Transferability with Checkpoints of a Single Model's Training (CVPR 2025 · similarity 0.88)
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.
