Research map / Adversarial robustness and security / Black-box and transfer attacks
Towards Transferable Targeted Attack
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 |
|---|---|---|---|
| replace cross-entropy with Poincaré distance (ablation) | iterative targeted attack · ImageNet-compatible, hold-out models | targeted attack success rate (%): 30.6 -> 35.9 | Helped |
| add triplet loss to Poincaré attack (ablation) | iterative targeted attack · ImageNet-compatible, hold-out models | targeted attack success rate (%): 35.9 -> 38.0 | Helped |
| add translation-invariant attack to Poincaré attack with triplet loss (ablation) | iterative targeted attack · ImageNet-compatible, hold-out models | targeted attack success rate (%): 38.0 -> 39.4 | Helped |
| use Poincaré + triplet + TI instead of DI²-FGSM | targeted attack · ImageNet-compatible, hold-out models | targeted attack success rate (%): 29.1 -> 39.4 | Helped |
| use Poincaré + triplet + TI instead of TI-FGSM | targeted attack · ImageNet-compatible, ensemble models | targeted attack success rate (%): 80.4 -> 86.3 | Helped |
| add triplet loss to Poincaré + TI (ablation) | attack on adversarially trained ensemble models · ImageNet-compatible, adversarially trained models | targeted attack success rate (%): 89.0 -> 91.9; 90.8 -> 92.1; 91.0 -> 91.4 | Mixed |
| add triplet loss to Poincaré + TI (ablation) | attack on adversarially trained hold-out models · ImageNet-compatible, adversarially trained hold-out models | targeted attack success rate (%): 17.7 -> 18.1; 12.9 -> 14.6; 7.7 -> 8.4 | Helped |
| use Poincaré distance rather than cross-entropy (ablation) | iterative targeted attack · ImageNet-compatible, iterative gradient updates | gradient magnitude | Helped |
| add triplet loss to Poincaré attack (ablation) | targeted adversarial examples · ImageNet-compatible, hold-out models | correctly classified adversarial examples | Helped |
Closest papers by meaning
- Towards Transferable Targeted Adversarial Examples (CVPR 2023 · similarity 0.95)
- Delving into Transferable Adversarial Examples and Black-box Attacks (ICLR 2017 · similarity 0.94)
- Boosting Adversarial Transferability through Augmentation in Hypothesis Space (CVPR 2025 · similarity 0.90)
- AIM: Additional Image Guided Generation of Transferable Adversarial Attacks (AAAI 2025 · similarity 0.89)
- Enhancing Adversarial Transferability with Adversarial Weight Tuning (AAAI 2025 · similarity 0.88)
- Everywhere Attack: Attacking Locally and Globally to Boost Targeted Transferability (AAAI 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.