Research map / Adversarial robustness and security / Attacks on text and graphs
Out of Sight, Out of Track: Adversarial Attacks on Propagation-based Multi-Object Trackers via Query State Manipulation
Halima Bouzidi, Haoyu Liu, Yonatan Achamyeleh, Praneetsai Iddamsetty, Mohammad Al Faruque
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 Cost and Siphon losses to flooding loss in TQF (ablation) | FADE-TQF on CO-MOT · MOT17 | HOTA under attack: 40.25 -> 37.26 | Improved |
| add L_Cost to TQF (ablation) | FADE-TQF on Samba · MOT17 | HOTA under attack: 53.14 -> 45.53 | Improved |
| add L_Siphon to TQF (ablation) | FADE-TQF on Samba · MOT17 | HOTA under attack: 44.96 -> 45.53 | Hurt |
| add L_Siphon to TQF (ablation) | FADE-TQF on CO-MOT · MOT17 | HOTA under attack: 39.00 -> 37.26 | Improved |
| add L_Decorr to TMC (ablation) | FADE-TMC on Samba · MOT17 | HOTA under attack: 61.63 -> 48.04 | Improved |
| add L_Decorr to TMC (ablation) | FADE-TMC on CO-MOT · MOT17 | HOTA under attack: 53.27 -> 41.73 | Improved |
| add L_Erase to TMC (ablation) | FADE-TMC on Samba · MOT17 | HOTA under attack: 52.31 -> 48.04 | Improved |
| add L_Erase to TMC (ablation) | FADE-TMC on CO-MOT and MeMOTR · MOT17 | HOTA under attack: CO-MOT 41.91 -> 41.73; MeMOTR 41.90 -> 41.56 | No clear effect |
| use TQF/TMC instead of association-focused Hijacking | MOTRv2 · MOT20 (digital) | HOTA under attack: 58.27 -> 29.64 | Improved |
| use FADE-TMC instead of Daedalus flooding | MeMOTR · MOT20 (digital) | HOTA under attack: 62.51 -> 37.70 | Improved |
| FADE-TQF versus FADE-TMC on MOTRv2 (MOT20) | FADE · MOT20 (digital) | HOTA under attack: TMC 56.68 vs TQF 29.64 | Mixed |
| increase PGD iterations from 5 to 100 for AAI attack (ablation) | FADE-TMC AAI on MOTR · MOT17 samples | HOTA under attack: 49.51 -> 39.85 | Improved |
Closest papers by meaning
- SPARK: Spatial-aware Online Incremental Attack Against Visual Tracking (ECCV 2020 · similarity 0.88)
- IoU Attack: Towards Temporally Coherent Black-Box Adversarial Attack for Visual Object Tracking (CVPR 2021 · similarity 0.87)
- Efficient Adversarial Attacks for Visual Object Tracking (ECCV 2020 · similarity 0.87)
- F&F Attack: Adversarial Attack against Multiple Object Trackers by Inducing False Negatives and False Positives (ICCV 2023 · similarity 0.87)
- LRR: Language-Driven Resamplable Continuous Representation against Adversarial Tracking Attacks (ICLR 2024 · similarity 0.84)
- Two-stream Beats One-stream: Asymmetric Siamese Network for Efficient Visual Tracking (AAAI 2025 · similarity 0.83)
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.
