Research map / Adversarial robustness and security / Black-box and transfer attacks
Evading Defenses to Transferable Adversarial Examples by Translation-Invariant Attacks
Check your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 translation-invariant gradient convolution to DIM | Ensemble of Inc-v3, Inc-v4, IncRes-v2, and Res-v2-152 · Inc-v3_ens3 | success rate (%): 66.0 -> 84.8 | Helped |
| Add translation-invariant gradient convolution to DIM | Ensemble of Inc-v3, Inc-v4, IncRes-v2, and Res-v2-152 · Inc-v3_ens4 | success rate (%): 63.3 -> 82.7 | Helped |
| Add translation-invariant gradient convolution to DIM | Ensemble of Inc-v3, Inc-v4, IncRes-v2, and Res-v2-152 · IncRes-v2_ens | success rate (%): 45.9 -> 78.0 | Helped |
| Add translation-invariant gradient convolution to DIM | Ensemble of Inc-v3, Inc-v4, IncRes-v2, and Res-v2-152 · HGD | success rate (%): 57.7 -> 82.6 | Helped |
| Add translation-invariant gradient convolution to DIM | Ensemble of Inc-v3, Inc-v4, IncRes-v2, and Res-v2-152 · R&P | success rate (%): 51.7 -> 81.4 | Helped |
| Add translation-invariant gradient convolution to DIM | Ensemble of Inc-v3, Inc-v4, IncRes-v2, and Res-v2-152 · JPEG | success rate (%): 82.5 -> 83.4 | Helped |
| Add translation-invariant gradient convolution to DIM | Ensemble of Inc-v3, Inc-v4, IncRes-v2, and Res-v2-152 · TVM | success rate (%): 64.1 -> 79.8 | Helped |
| Add translation-invariant gradient convolution to DIM | Ensemble of Inc-v3, Inc-v4, IncRes-v2, and Res-v2-152 · NIPS-r3 | success rate (%): 63.7 -> 83.1 | Helped |
| Replace the Gaussian kernel with a uniform kernel (ablation) | TI-DIM · Inc-v3_ens3 | success rate (%): 46.9 -> 32.6 | Hurt |
| Replace the Gaussian kernel with a linear kernel (ablation) | TI-DIM · Inc-v3_ens3 | success rate (%): 46.9 -> 45.2 | No clear effect |
| Use a linear rather than uniform kernel (ablation) | TI-FGSM · Inc-v3_ens3 | success rate (%): 25.0 -> 30.7 | Helped |
| Use a Gaussian rather than uniform kernel (ablation) | TI-DIM · Inc-v3_ens3 | success rate (%): 32.6 -> 46.9 | Helped |
Closest papers by meaning
- Improving Transferability of Adversarial Examples With Input Diversity (CVPR 2019 · similarity 0.93)
- Delving into Transferable Adversarial Examples and Black-box Attacks (ICLR 2017 · similarity 0.92)
- Cross-Domain Transferability of Adversarial Perturbations (NeurIPS 2019 · similarity 0.92)
- Boosting Adversarial Transferability through Augmentation in Hypothesis Space (CVPR 2025 · similarity 0.91)
- Rethinking Model Ensemble in Transfer-based Adversarial Attacks (ICLR 2024 · similarity 0.90)
- Enhancing Adversarial Transferability with Adversarial Weight Tuning (AAAI 2025 · similarity 0.90)
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.