Research map / Adversarial robustness and security / Attacks on text and graphs
TkML-AP: Adversarial Attacks to Top-k Multi-Label Learning
Shu Hu, Lipeng Ke, Xin Wang, Siwei Lyu
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 |
|---|---|---|---|
| use TkML-AP-U instead of kFool | untargeted attack · MS COCO 2014 top-3 | ASR (%): 61.4 -> 100 | Improved |
| use TkML-AP-U instead of kFool | untargeted attack · PASCAL VOC 2012 top-10 | ASR (%): 88.7 -> 98.3 | Improved |
| use TkML-AP-U instead of kFool (per-pixel perturbation) | untargeted attack · MS COCO 2014 top-10 | Pert (x10^-2): 16.44 -> 0.59 | Improved |
| attack with k=3 but system uses larger k'=5 | TkML-AP-U · PASCAL VOC 2012 | ASR (%): 99.6 -> 3.6 | Hurt |
| attack with k=3 but system uses larger k'=10 | TkML-AP-U · MS COCO 2014 | ASR (%): 100 -> 3.9 | Hurt |
| use TkML-AP-Uv instead of kUAPs | universal untargeted attack · MS COCO 2014 top-1 | ASR (%): 63.9 -> 86.5 | Improved |
| use TkML-AP-Uv instead of kUAPs (perturbation) | universal untargeted attack · MS COCO 2014 top-1 | Pert: 0.51 -> 0.13 | Improved |
| use TkML-AP-Uv instead of kUAPs | universal untargeted attack · MS COCO 2014 top-3 | ASR (%): 73.2 -> 80.5 | Improved |
| use TkML-AP-T instead of ML-AP | targeted attack Worst case · PASCAL VOC 2012 top-10 | ASR (%): 39.1 -> 57 | Improved |
| use TkML-AP-T instead of ML-AP | targeted attack Worst case · MS COCO 2014 top-10 | ASR (%): 59 -> 73.1 | Improved |
| use TkML-AP-T instead of ML-AP | targeted attack Random case · PASCAL VOC 2012 top-10 | ASR (%): 67.7 -> 76.4 | Improved |
| use TkML-AP-T instead of ML-AP | targeted attack Best case · MS COCO 2014 top-3 | ASR (%): 100 -> 100 | No clear effect |
Closest papers by meaning
- GSBA$^K$: $top$-$K$ Geometric Score-based Black-box Attack (ICLR 2025 · similarity 0.87)
- Semantic-Aware Multi-Label Adversarial Attacks (CVPR 2024 · similarity 0.85)
- Adversarial Perturbations Are Formed by Iteratively Learning Linear Combinations of the Right Singular Vectors of the Adversarial Jacobian (ICML 2025 · similarity 0.84)
- Soften to Defend: Towards Adversarial Robustness via Self-Guided Label Refinement (CVPR 2024 · similarity 0.83)
- LimeAttack: Local Explainable Method for Textual Hard-Label Adversarial Attack (AAAI 2024 · similarity 0.83)
- PatchDEMUX: A Certifiably Robust Framework for Multi-label Classifiers Against Adversarial Patches (CVPR 2025 · similarity 0.82)
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.
