Research map / Adversarial robustness and security / Backdoor and poisoning
BadTrack: A Poison-Only Backdoor Attack on Visual Object Tracking
Bin Huang, Jiaqian Yu, Yiwei Chen, Siyang Pan, Qiang Wang, Zhi Wang
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 clean-label BadTrack poisoning | SiamRPN++ · LaSOT poisoned set | AUC (%): 44.83 -> 19.51 | Hurt |
| add dirty-label BadTrack poisoning | SiamRPN++ · LaSOT poisoned set | AUC (%): 44.83 -> 7.24 | Hurt |
| add dirty-label BadTrack poisoning | OSTrack (with CE) · LaSOT poisoned set | AUC (%): 67.48 -> 67.78 | No clear effect |
| add clean-label BadTrack poisoning | OSTrack (with CE) · LaSOT poisoned set | AUC (%): 67.48 -> 17.49 | Hurt |
| add clean-label BadTrack poisoning, no candidate elimination modules (ablation) | OSTrack (w/o CE) · LaSOT poisoned set | AUC (%): 68.19 -> 20.28 | Hurt |
| use more complex trigger: pure white, chess board, colorful (ablation) | clean-label BadTrack on SiamRPN++ · LaSOT poisoned set | AUC (%): 28.05 -> 19.51 -> 9.13 | Hurt |
| increase attack ratio alpha (ablation) | clean-label BadTrack on SiamRPN++ · LaSOT | AUC / Pr / Pnorm | Mixed |
| compare dirty-label BadTrack with FSBA on clean set | SiamRPN++ · GOT10k clean set of backdoored tracker | AO (%): 54.20 -> 66.26 | Improved |
| compare dirty-label BadTrack with FSBA on poisoned set | SiamRPN++ · LaSOT poisoned set | AUC (%): 5.61 -> 7.24 | No clear effect |
| apply hue, brightness, saturation, contrast jittering as defense | clean-label BadTrack on SiamRPN++ · VOT2018 | EAO / Acc | No clear effect |
| fine-tune on 5% clean training data for 20 epochs as defense | clean-label BadTrack on SiamRPN++ · VOT2018 poisoned set | EAO: about 0.2 | No clear effect |
| add Gaussian noise or head pruning as defense | clean-label BadTrack on OSTrack/SiamRPN++ · VOT2018 / LaSOT | EAO / AUC | No clear effect |
Closest papers by meaning
- Few-Shot Backdoor Attacks on Visual Object Tracking (ICLR 2022 · similarity 0.93)
- IoU Attack: Towards Temporally Coherent Black-Box Adversarial Attack for Visual Object Tracking (CVPR 2021 · similarity 0.86)
- Defending Backdoor Attacks on Vision Transformer via Patch Processing (AAAI 2023 · similarity 0.86)
- Test-Time Backdoor Detection for Object Detection Models (CVPR 2025 · similarity 0.85)
- Attention-Imperceptible Backdoor Attacks on Vision Transformers (AAAI 2025 · similarity 0.85)
- Not All Prompts Are Secure: A Switchable Backdoor Attack Against Pre-trained Vision Transfomers (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.
