Research map / Adversarial robustness and security / Black-box and transfer attacks
Adaptive Cross-Modal Transferable Adversarial Attacks From Images to Videos
Zhipeng Wei, Jingjing Chen, Zuxuan Wu, Yu–Gang Jiang
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 |
|---|---|---|---|
| replace DR feature-variance objective with cosine-similarity minimization (I2V) | image-to-video transfer attack · UCF-101 and Kinetics-400 | AASR: boost of more than 63.33% and 42.51% | Improved |
| use Resnet-101 instead of Alexnet as the white-box image model (ablation) | I2V · UCF-101, NL-101 target | ASR (%): 50.49 -> 71.28 | Improved |
| use Resnet-101 instead of Squeezenet as the white-box image model (ablation) | I2V · Kinetics-400, TPN-101 target | ASR (%): 55.50 -> 87.25 | Improved |
| attack an ensemble of image models (ENS-I2V) instead of single Resnet-101 | I2V · UCF-101 | AASR (%): to 65.68% | Improved |
| attack an ensemble of image models (ENS-I2V) | I2V · Kinetics-400 | AASR (%): to 77.88% | Improved |
| ensemble of image models instead of Resnet-101 alone (ablation) | I2V · UCF-101, TPN-101 target | ASR (%): 61.38 -> 52.47 | Hurt |
| ensemble of image models instead of Resnet-101 alone (ablation) | I2V · Kinetics-400, TPN-50 target | ASR (%): 90.25 -> 88.00 | Hurt |
| perturb middle layers (layer 2 or 3) instead of bottom or top layers of image models (ablation) | I2V · Kinetics-400 | AASR | Improved |
| use larger step size with fewer iterations (ablation) | I2V · Kinetics-400 | AASR | Improved |
| ENS-I2V (no white-box video model) versus ILAF-fine-tuned transfer attacks using NL-101 or SlowFast-101 as white-box | ENS-I2V · UCF-101 and Kinetics-400 | AASR | Improved |
| I2V versus transfer baselines using TPN-101 as white-box model | I2V · Kinetics-400 | AASR | Hurt |
Closest papers by meaning
- Cross-Modal Transferable Adversarial Attacks From Images to Videos (CVPR 2022 · similarity 0.99)
- Global-Local Characteristic Excited Cross-Modal Attacks from Images to Videos (AAAI 2023 · similarity 0.94)
- Transferability of Adversarial Attacks in Video-based MLLMs: A Cross-modal Image-to-Video Approach (AAAI 2026 · similarity 0.91)
- Boosting the Transferability of Video Adversarial Examples via Temporal Translation (AAAI 2022 · similarity 0.90)
- Enhancing Cross-Task Black-Box Transferability of Adversarial Examples With Dispersion Reduction (CVPR 2020 · similarity 0.90)
- Enhancing Adversarial Transferability with Adversarial Weight Tuning (AAAI 2025 · similarity 0.88)
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.
