Research map / Adversarial robustness and security / Attacks on text and graphs
Dual Attention Suppression Attack: Generate Adversarial Camouflage in Physical World
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 DAS instead of MeshAdv | ResNet-152 vehicle classification in the digital world · Digital vehicle classification | Accuracy (%): 35.33 -> 32.49 | Helped |
| Use DAS instead of CAMOU | Faster R-CNN vehicle detection in the digital world · Digital vehicle detection | P@0.5 (%): 69.64 -> 62.11 | Helped |
| Use DAS instead of CAMOU | ResNet-152 vehicle classification in the physical world · Physical vehicle classification | Accuracy (%): 31.25 -> 29.86 | Helped |
| Use DAS instead of UPC | Faster R-CNN vehicle detection in the physical world · Physical vehicle detection | P@0.5 (%): 52.08 -> 44.44 | Helped |
| Add the attention distraction loss L_d rather than using only L_e (ablation) | DAS · Digital vehicle classification, Inception-V3 | Accuracy (%): 59.87 -> 36.53 | Helped |
| Add the human attention evasion loss L_e to L_d (ablation) | DAS · Digital vehicle classification, Inception-V3 | Accuracy (%): 36.53 -> 39.86 | Hurt |
| Use only the human attention evasion loss L_e instead of combined L_d + λL_e (ablation) | DAS · Digital vehicle classification, Inception-V3 | Accuracy (%): 39.86 -> 59.87 | Hurt |
| Use only L_e rather than L_d (ablation) | DAS · Digital vehicle classification across VGG-19 and DenseNet versus Inception-V3 and ResNet-152 | Accuracy (%) | Mixed |
| Increase λ (ablation) | DAS · Vehicle classification, four-model average | Accuracy | Hurt |
| Increase λ (ablation) | DAS · Vehicle camouflage appearance | SSIM: 0.7034, 0.7551, 0.8750, 0.9982, 0.9991, and 0.9998 | Helped |
Closest papers by meaning
- Adversarial Camouflage: Hiding Physical-World Attacks With Natural Styles (CVPR 2020 · similarity 0.91)
- Unnoticed Yet Effective: A Hybrid Physical Camouflage Framework Against DNNs and Human Perception (AAAI 2026 · similarity 0.91)
- PhyCamo: A Robust Physical Camouflage via Contrastive Learning for Multi-View Physical Adversarial Attack (AAAI 2025 · similarity 0.89)
- CNCA: Toward Customizable and Natural Generation of Adversarial Camouflage for Vehicle Detectors (NeurIPS 2024 · similarity 0.86)
- RAUCA: A Novel Physical Adversarial Attack on Vehicle Detectors via Robust and Accurate Camouflage Generation (ICML 2024 · similarity 0.85)
- Breaking Barriers in Physical-World Adversarial Examples: Improving Robustness and Transferability via Robust Feature (AAAI 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.