Research map / Adversarial robustness and security / Attacks on text and graphs
Adversarial Attention Perturbations for Large Object Detection Transformers
Zachary Yahn, Selim Furkan Tekin, Fatih Ilhan, Sihao Hu, Tiansheng Huang, Yichang Xu, Margaret Loper, Ling Liu
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 |
|---|---|---|---|
| add the learnable adversarial attention map (ablation) | AFOG · 12 detection transformers on COCO | adversarial mAP reduction: up to 30.6% on InternImage, average 15.1% | Improved |
| set attention learning rate to 0.1 versus zero or other values (ablation) | AFOG · DETR-type models on COCO | attack improvement: 0.1 strongest | Improved |
| use AFOG-V (victim-loss variant) instead of generic AFOG | AFOG · COCO, Deformable-DETR | adversarial mAP: 4.8 -> 1.5 | Improved |
| use AFOG-V instead of generic AFOG | AFOG · COCO, DETR-R50 | adversarial mAP: 4.1 -> 4.5 | Hurt |
| use AFOG-F instead of generic AFOG | AFOG · COCO, AlignDETR | adversarial mAP: 18.1 -> 1.4 | Improved |
| use AFOG-F instead of generic AFOG | AFOG · COCO, DETR-R50 | adversarial mAP: 4.1 -> 9.8 | Hurt |
| AFOG versus best prior attack on Swin | AFOG · COCO, Swin | adversarial mAP: 42.3 -> 7.3 | Improved |
| AFOG versus TOG on SSD-300 | AFOG · VOC07, SSD-300 | adversarial mAP: 0.86 -> 0.50 | Improved |
| AFOG versus TOG on YOLOv3 | AFOG · VOC07, YOLOv3 | adversarial mAP: 0.56 -> 2.28 | Hurt |
| AFOG versus best CNN-attack DAG-free TOG on Faster R-CNN | AFOG · VOC07, Faster R-CNN | adversarial mAP: 2.64 -> 2.38 | Improved |
Closest papers by meaning
- Transferable Adversarial Attacks for Object Detection Using Object-Aware Significant Feature Distortion (AAAI 2024 · similarity 0.87)
- Adversarial Examples for Semantic Segmentation and Object Detection (ICCV 2017 · similarity 0.86)
- Towards Adversarially Robust Object Detection (ICCV 2019 · similarity 0.86)
- DESTR: Object Detection With Split Transformer (CVPR 2022 · similarity 0.85)
- ObjectAdv: Object-Level Unrestricted Adversarial Attacks via Diffusion Models (AAAI 2026 · similarity 0.85)
- Towards Understanding and Improving Adversarial Robustness of Vision Transformers (CVPR 2024 · 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.
