Research map / Adversarial robustness and security / Black-box and transfer attacks
S 4 ST: A Strong, Self-Transferable, faSt, and Simple Scale Transformation for Data-Free Transferable Targeted Attack
Yongxiang Liu, Bowen Peng, Li Liu, Xiang Li
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 S4ST-Aug (complementary augmentation pool) (ablation) | S4ST-Base · ResNet-50 surrogate | average tSuc: 58.2 -> 76.8 | Improved |
| add S4ST-Block (block-wise scaling, m=6) (ablation) | S4ST-Base · ResNet-50 surrogate | average tSuc: 58.2 -> 74.3 | Improved |
| combine Aug and Block (ablation) | S4ST-Base · ResNet-50 surrogate | average tSuc: 58.2 -> 83.0 | Improved |
| add flip to augmentation pool (ablation) | S4ST · ResNet-50 surrogate | average tSuc: 80.8 -> 83.0 | Improved |
| add contrast to augmentation pool (ablation) | S4ST · ResNet-50 surrogate | average tSuc: 82.6 -> 83.0 | No clear effect |
| add brightness to augmentation pool (ablation) | S4ST · ResNet-50 surrogate | average tSuc: 82.7 -> 83.0 | No clear effect |
| add saturation to augmentation pool (ablation) | S4ST · ResNet-50 surrogate | average tSuc: 82.7 -> 83.0 | No clear effect |
| add hue to augmentation pool (ablation) | S4ST · ResNet-50 surrogate | average tSuc: 82.6 -> 83.0 | No clear effect |
| use scaling instead of shear as the basic transformation | TMI · ResNet-50 surrogate | average tSuc: 48.8% -> 58.0% | Improved |
| tune DI and RDI parameters | DI / RDI · ResNet-50 surrogate | average tSuc: 22.4% -> 36.6% (DI), 27.3% -> 41.7% (RDI) | Improved |
| use bidirectional (upscale and downscale) scaling | DI/RDI · ResNet-50 surrogate | average tSuc: 41.7% -> 56.8% | Improved |
| replace simple bidirectional scaling with S4ST-Base | scaling · ResNet-50 surrogate | average tSuc: 57.2% -> 61.6% | Improved |
Closest papers by meaning
- Everywhere Attack: Attacking Locally and Globally to Boost Targeted Transferability (AAAI 2025 · similarity 0.86)
- Boosting the Transferability of Adversarial Attack on Vision Transformer with Adaptive Token Tuning (NeurIPS 2024 · similarity 0.85)
- Your Scale Factors are My Weapon: Targeted Bit-Flip Attacks on Vision Transformers via Scale Factor Manipulation (CVPR 2025 · similarity 0.85)
- PubDef: Defending Against Transfer Attacks From Public Models (ICLR 2024 · similarity 0.84)
- Improving Transferable Targeted Adversarial Attacks with Model Self-Enhancement (CVPR 2024 · similarity 0.84)
- Harnessing the Computation Redundancy in ViTs to Boost Adversarial Transferability (NeurIPS 2025 · similarity 0.84)
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.
