Research map / Adversarial robustness and security / Certified and empirical robustness
Transferable, Controllable, and Inconspicuous Adversarial Attacks on Person Re-identification With Deep Mis-Ranking
Hongjun Wang, Guangrun Wang, Ya Li, Dongyu Zhang, Liang Lin
Check 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 |
|---|---|---|---|
| Apply the proposed attack to AlignedReID | AlignedReID · Market1501 | Rank-1: 91.8 -> 1.4 | Improved |
| Use mis-ranking loss instead of modified misclassification loss alone (ablation) | AlignedReID attack · Market1501 | Rank-1: 13.7 -> 4.5 | Improved |
| Combine modified xent with mis-ranking loss (ablation) | AlignedReID attack · Market1501 | Rank-1: 4.5 -> 1.4 | Improved |
| Replace conventional misclassification loss with the proposed combined losses (ablation) | AlignedReID attack · Market1501 | Rank-1: 28.5 -> 1.4 | Improved |
| Replace PatchGAN with a multi-stage discriminator (ablation) | AlignedReID attack · Market1501, ε=40 | Rank-1: 48.3 -> 0.0 | Improved |
| Replace PatchGAN with a multi-stage discriminator (ablation) | AlignedReID attack · Market1501, ε=10 | Rank-1: 53.3 -> 24.4 | Improved |
| Reduce perturbation bound from ε=16 to ε=10 (ablation) | AlignedReID attack · Market1501 | Rank-1: 1.4 -> 24.4 | Hurt |
| Constrain attack locations to pixel ratio 1/2 rather than full size (ablation) | AlignedReID attack · Market1501 | Rank-1: 1.4 -> 39.3 | Hurt |
| Constrain attack locations to pixel ratio 1/4 without increased magnitude (ablation) | AlignedReID attack · Market1501 | Rank-1: 1.4 -> 72.7 | Hurt |
| Increase noise magnitude at pixel ratio 1/4 (ablation) | AlignedReID attack · Market1501 | Rank-1: 72.7 -> 0.3 | Improved |
| Use learned attack locations instead of random locations (ablation) | AlignedReID attack with Gaussian noise · Market1501 | Rank-1: 81.9 -> 79.4 | Improved |
| Use learned attack locations instead of random locations (ablation) | AlignedReID attack with uniform noise · Market1501 | Rank-1: 51.1 -> 50.7 | Improved |
Closest papers by meaning
- Multi-Expert Adversarial Attack Detection in Person Re-Identification Using Context Inconsistency (ICCV 2021 · similarity 0.91)
- Debiased Dual-Invariant Defense for Adversarially Robust Person Re-Identification (AAAI 2026 · similarity 0.90)
- Pose Transferrable Person Re-Identification (CVPR 2018 · similarity 0.88)
- Multiform Attack for Transferable Cross-Modal Person Re-Identification (NeurIPS 2026 · similarity 0.88)
- Feature-Level Adversarial Attacks and Ranking Disruption for Visible-Infrared Person Re-identification (NeurIPS 2024 · similarity 0.86)
- Exploring the Camera Bias of Person Re-identification (ICLR 2025 · similarity 0.85)
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.
