Research map / Adversarial robustness and security / Backdoor and poisoning
Few-Shot Backdoor Attacks on Visual Object Tracking
Yiming Li, Haoxiang Zhong, Xingjun Ma, Yong Jiang, Shu-Tao Xia
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 |
|---|---|---|---|
| Use feature-space loss maximization alongside tracking-loss minimization instead of BOBA label flipping | SiamFC++ · OTB100, one-shot | AUC-A: 57.79 -> 18.78 | Improved |
| Use feature-space loss maximization alongside tracking-loss minimization instead of BOBA label flipping | SiamFC++ · GOT10K validation, one-shot | AUC-A: 66.70 -> 35.82 | Improved |
| Use feature-space loss maximization alongside tracking-loss minimization instead of BOBA label flipping | SiamFC++ · LaSOT, one-shot | AUC-A: 43.06 -> 16.39 | Improved |
| Increase the poisoning rate from 0% to 10% (ablation) | FSBA on SiamFC · OTB100, one-shot | Pr-A: 72.43 -> 11.06 | Improved |
| Increase the poisoning rate from 0% to 10% (ablation) | FSBA on SiamFC · OTB100, few-shot | Pr-A: 74.03 -> 7.92 | Improved |
| Increase the poisoning rate from 0% to 10% (ablation) | FSBA on SiamFC · OTB100 | Pr-B: 79.23 -> 75.98 | Hurt |
| Train SiamFC++ with additional GOT10K data (ablation) | FSBA · OTB100, one-shot | AUC-A: 18.78 -> 23.20 | Hurt |
| Train SiamFC++ with additional LaSOT data (ablation) | FSBA · OTB100, one-shot | AUC-A: 23.20 -> 23.19 | No clear effect |
| Apply frame-wise color jittering as a defense (ablation) | FSBA SiamFC++ · OTB100, one-shot | Pr-A: Pr-A is still below 40% in all cases | No clear effect |
| Fine-tune using benign samples (ablation) | FSBA SiamFC++ · OTB100, one-shot | Pr-A: performance gap ... is still larger than 40% in all cases | No clear effect |
| Add Gaussian noise to test frames (ablation) | FSBA SiamFC++ · OTB100, one-shot | Pr-A: <30% | No clear effect |
| Prune model channels at rates from 0% to 30% (ablation) | FSBA SiamFC++ · OTB100, one-shot | Pr-A: 25.52; 26.70; 27.40; 27.42; 29.04; 25.29; 21.69 | No clear effect |
Closest papers by meaning
- BadTrack: A Poison-Only Backdoor Attack on Visual Object Tracking (NeurIPS 2023 · similarity 0.93)
- Fooling Detection Alone is Not Enough: Adversarial Attack against Multiple Object Tracking (ICLR 2020 · similarity 0.87)
- IoU Attack: Towards Temporally Coherent Black-Box Adversarial Attack for Visual Object Tracking (CVPR 2021 · similarity 0.87)
- Attention-Imperceptible Backdoor Attacks on Vision Transformers (AAAI 2025 · similarity 0.86)
- Not All Prompts Are Secure: A Switchable Backdoor Attack Against Pre-trained Vision Transfomers (CVPR 2024 · similarity 0.85)
- BadVLA: Towards Backdoor Attacks on Vision-Language-Action Models via Objective-Decoupled Optimization (NeurIPS 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.
