Research map / Adversarial robustness and security / Black-box and transfer attacks
Efficient Robustness Assessment via Adversarial Spatial-Temporal Focus on Videos
Xingxing Wei, Songping Wang, Huanqian Yan
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 spatial agent (key regions) (ablation) | PGD+NES baseline · HMDB-51 val, C3D | fooling rate: 73.3 -> 88.3 | Improved |
| add temporal agent (key frames) (ablation) | PGD+NES baseline · HMDB-51 val, C3D | fooling rate: 73.3 -> 93.3 | Improved |
| add both agents jointly (ablation) | PGD+NES baseline · HMDB-51 val, C3D | query number: 3662 -> 2227 | Improved |
| add both agents jointly (ablation) | PGD+NES baseline · HMDB-51 val, C3D | MAP: 6.37 -> 3.35 | Improved |
| add edgebox reward (ablation) | common reward only · HMDB-51 val, C3D | query number: 3780 -> 2690 | Improved |
| add sparse reward (ablation) | common + edgebox · HMDB-51 val, C3D | fooling rate: 91.7 -> 96.7 | Improved |
| add representative reward (ablation) | common + edgebox + sparse · HMDB-51 val, C3D | query number: 2490 -> 2227 | Improved |
| all rewards vs common reward only (ablation) | AstFocus · HMDB-51 val, C3D | fooling rate: 76.7 -> 100 | Improved |
| replace random agents with RL-trained agents (ablation) | AstFocus · HMDB-51 val, C3D | fooling rate: 86.7 -> 100 | Improved |
| replace random agents with RL-trained agents (ablation) | AstFocus · HMDB-51 val, C3D | query number: 2632 -> 2227 | Improved |
| replace NES with Prior convictions or ZOO gradient estimator | AstFocus · HMDB-51 | FR, QN, MAP | No clear effect |
| apply PGD-AT defense to target model | AstFocus attack · HMDB-51, C3D | fooling rate: 92.0 -> 60.0 | Hurt |
Closest papers by meaning
- Sparse Black-Box Video Attack with Reinforcement Learning (IJCV 2022 · similarity 0.90)
- Heuristic Black-Box Adversarial Attacks on Video Recognition Models (AAAI 2020 · similarity 0.88)
- From Pretrain to Pain: Adversarial Vulnerability of Video Foundation Models Without Task Knowledge (AAAI 2026 · similarity 0.87)
- Reinforcement Learning Platform for Adversarial Black-box Attacks with Custom Distortion Filters (AAAI 2025 · similarity 0.85)
- FrameShield: Adversarially Robust Video Anomaly Detection (NeurIPS 2025 · similarity 0.84)
- Robust SAM: On the Adversarial Robustness of Vision Foundation Models (AAAI 2025 · 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.
