Research map / Adversarial robustness and security / Black-box and transfer attacks
BASAR:Black-Box Attack on Skeletal Action Recognition
Yunfeng Diao, Tianjia Shao, Yong-Liang Yang, Kun Zhou, He Wang
Check 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 BASAR instead of the strongest reported baseline | skeletal action recognizers · HDM05 and NTU targeted attack | success rate: 30.31% -> 100.00% | Improved |
| use BASAR instead of SMART (HRNN) | skeletal action recognizers · HDM05 untargeted attack | success rate: 89.25% -> 100.00% | Improved |
| use BASAR instead of SMART (HRNN) | skeletal action recognizers · NTU untargeted attack | success rate: 99.33% -> 100.00% | Improved |
| add manifold projection (ablation) | BASAR, HDM05 untargeted on STGCN · HDM05 | on-manifold sample percentage: 0.00% -> 99.55% | Improved |
| add manifold projection (ablation) | BASAR, HDM05 untargeted on STGCN · HDM05 | joint angular acceleration deviation: 0.34 -> 0.11 | Improved |
| add manifold projection (ablation) | BASAR, HDM05 untargeted on STGCN · HDM05 | bone-length deviation: 0.66% -> 0.00% | Improved |
| add manifold projection (ablation) | BASAR, HDM05 untargeted on STGCN · HDM05 | joint position deviation: 0.10 -> 0.13 | Hurt |
| add manifold projection (ablation) | BASAR, HDM05 targeted on STGCN · HDM05 | on-manifold sample percentage: 0.00% -> 56.98% | Improved |
| add manifold projection (ablation) | BASAR · HDM05 perceptual naturalness | naturalness preference | Improved |
| use semantic joint features | SGN compared with STGCN and MSG3D · HDM05, NTU, and Kinetics attacks | perturbation required to fool the classifier | Improved |
| use explicit bone information | MSG3D compared with STGCN · skeletal action recognition attacks | robustness to attack | Improved |
| apply BASAR rather than the white-box attack in the cited work | HDM05 perceptual indistinguishability study · HDM05 | indistinguishability: 80.83% -> 89.90% | Improved |
Closest papers by meaning
- Hard No-Box Adversarial Attack on Skeleton-Based Human Action Recognition with Skeleton-Motion-Informed Gradient (ICCV 2023 · similarity 0.92)
- TASAR: Transfer-based Attack on Skeletal Action Recognition (ICLR 2025 · similarity 0.90)
- MaskCLR: Attention-Guided Contrastive Learning for Robust Action Representation Learning (CVPR 2024 · similarity 0.85)
- Revealing Key Details to See Differences: A Novel Prototypical Perspective for Skeleton-based Action Recognition (CVPR 2025 · similarity 0.85)
- Neuron: Learning Context-Aware Evolving Representations for Zero-Shot Skeleton Action Recognition (CVPR 2025 · similarity 0.85)
- Skeleton-based Action Recognition with Non-linear Dependency Modeling and Hilbert-Schmidt Independence Criterion (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.
