Research map / Adversarial robustness and security / Black-box and transfer attacks
ColorFool: Semantic Adversarial Colorization
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 |
|---|---|---|---|
| Use semantic-region color priors rather than unrestricted color changes (ablation) | ColorFool · CIFAR-10, R18 attack → R18 test | success rate: .954 -> .971 | Helped |
| Use semantic-region color priors rather than unrestricted color changes (ablation) | ColorFool · CIFAR-10, R18 attack → AN test | success rate: .689 -> .696 | No clear effect |
| Use semantic-region color priors rather than unrestricted color changes (ablation) | ColorFool · P-Places365, R50 attack | NIMA: 5.24 -> 5.22 | Hurt |
| Use semantic-region color priors rather than unrestricted color changes (ablation) | ColorFool · ImageNet, R50 attack | NIMA: 5.24 -> 5.24 | No clear effect |
| Use semantic-region color priors rather than unrestricted color changes | ColorFool · CIFAR-10, R18 attack → AN test | success rate: .696 vs .705 | Hurt |
| Use unrestricted semantic color perturbations instead of restricted pixel perturbations | adversarial attacks · P-Places365, CIFAR-10, and ImageNet | success rate after image filtering: above 60% | Helped |
| Use unrestricted semantic color perturbations instead of restricted pixel perturbations | adversarial attacks · P-Places365 | undetectability: 5%, 19%, 1% and 11% vs 73%, 72% and 75% | Helped |
| Add ColorFool color perturbations against PCL and adversarial training | ColorFool · CIFAR-10, ResNet110 with PCL+AdvT | success rate: .691 -> .999 | Helped |
Closest papers by meaning
- Natural Color Fool: Towards Boosting Black-box Unrestricted Attacks (NeurIPS 2022 · similarity 0.91)
- Unrestricted Adversarial Examples via Semantic Manipulation (ICLR 2020 · similarity 0.88)
- Towards Large Yet Imperceptible Adversarial Image Perturbations With Perceptual Color Distance (CVPR 2020 · similarity 0.87)
- Constructing Semantics-Aware Adversarial Examples with a Probabilistic Perspective (NeurIPS 2024 · similarity 0.84)
- CosPGD: an efficient white-box adversarial attack for pixel-wise prediction tasks (ICML 2024 · similarity 0.84)
- A Frustratingly Simple Yet Highly Effective Attack Baseline: Over 90% Success Rate Against the Strong Black-box Models of GPT-4.5/4o/o1 (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.