Research map / Adversarial robustness and security / Attacks on text and graphs
Just One Moment: Structural Vulnerability of Deep Action Recognition Against One Frame Attack
Jaehui Hwang, Jun-Hyuk Kim, Jun-Ho Choi, Jong-Seok Lee
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 |
|---|---|---|---|
| increase epsilon from 2 to 16 | one frame attack on I3D · Kinetics-400 | fooling rate: 0.83 -> 1.00 | Improved |
| increase epsilon from 2 to 16 | one frame attack on SlowFast · Kinetics-400 | fooling rate: 0.73 -> 0.98 | Improved |
| increase epsilon from 2 to 16 | one frame attack on ir-CSN · Kinetics-400 | fooling rate: 0.60 -> 0.97 | Improved |
| use one frame attack on the most vulnerable frame instead of Wei et al. LSTM-targeted attack | I3D · Kinetics-400 | fooling rate: 0.81 -> 0.99 | Improved |
| use one frame attack instead of Wei et al. attack at epsilon 2 | ir-CSN · Kinetics-400 | fooling rate: 0.68 -> 0.60 | Hurt |
| perturb only structurally vulnerable frames (periodic indices) instead of other frames (ablation) | I3D and SlowFast · Kinetics-400 | fooling rate | Improved |
| increase number of videos used for universal perturbation (ablation) | universal one frame attack · Kinetics-400 | fooling rate | Improved |
| use multiple vulnerable frames instead of only the most vulnerable one for universal perturbation (ablation) | universal one frame attack on I3D and SlowFast · Kinetics-400 | fooling rate: above 80% -> exceed 90% | Improved |
| use time-invariant universal attack perturbing P=4 frames per 32 | I3D and SlowFast · Kinetics-400 | fooling rate | Improved |
Closest papers by meaning
- Sparse Adversarial Perturbations for Videos (AAAI 2019 · similarity 0.86)
- From Pretrain to Pain: Adversarial Vulnerability of Video Foundation Models Without Task Knowledge (AAAI 2026 · similarity 0.85)
- Understanding the Robustness of Skeleton-Based Action Recognition Under Adversarial Attack (CVPR 2021 · similarity 0.85)
- Attacking Video Recognition Models with Bullet-Screen Comments (AAAI 2022 · similarity 0.85)
- Temporal-Distributed Backdoor Attack against Video Based Action Recognition (AAAI 2024 · similarity 0.84)
- LogoStyleFool: Vitiating Video Recognition Systems via Logo Style Transfer (AAAI 2024 · 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.
