Research map / Adversarial robustness and security / Black-box and transfer attacks
Vulnerabilities in Video Quality Assessment Models: The Challenge of Adversarial Attacks
Aoxiang Zhang, Yu Ran, Weixuan Tang, Yuan-Gen Wang
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 |
|---|---|---|---|
| patch-based instead of pixel-based random search (SimBA) (ablation) | black-box attack on VSFA · KoNViD-1k | SRCC after attack: 0.7831 -> -0.0305 | Improved |
| patch-based instead of pixel-based random search (SimBA) (ablation) | black-box attack on VSFA · LSVQ-Test | SRCC after attack: 0.7747 -> 0.2386 | Improved |
| increase T from 1 to 8 (black-box) (ablation) | black-box attack on VSFA · KoNViD-1k | SRCC after attack: -0.0305 -> 0.2747 | Hurt |
| increase T from 1 to 8 (white-box) (ablation) | white-box attack on VSFA · KoNViD-1k | SRCC after attack: -0.7003 -> -0.6964 | No clear effect |
| add motion embeddings (ablation) | BVQA-2022 · KoNViD-1k | SRCC under black-box attack: -0.1779 -> 0.1978 | Improved |
| add motion embeddings (ablation) | VSFA · KoNViD-1k | SRCC under black-box attack: -0.0305 -> 0.1858 | Improved |
| add motion embeddings (ablation) | VSFA · LSVQ-Test | SRCC under black-box attack: 0.2386 -> 0.4358 | Improved |
| add motion embeddings (ablation) | BVQA-2022 · LSVQ-Test | SRCC under black-box attack: 0.1917 -> 0.3355 | Improved |
| perturb 1/10 of frames instead of all (white-box) (ablation) | white-box attack on VSFA · KoNViD-1k | SRCC after attack: -0.7198 -> -0.2692 | Mixed |
| perturb 1/10 of frames instead of all (black-box) (ablation) | black-box attack on VSFA · KoNViD-1k | SRCC after attack: -0.0305 -> 0.6836 | Hurt |
| Score-Reversed Boundary Loss instead of MSE loss (ablation) | black-box and white-box attack on VSFA · KoNViD-1k | query efficiency | Improved |
| L_infinity instead of L2 constraint in white-box attack (ablation) | white-box attack on NR-VQA models · five datasets | SRCC degradation: 194% vs 195% | No clear effect |
Closest papers by meaning
- Comparing the Robustness of Modern No-Reference Image- and Video-Quality Metrics to Adversarial Attacks (AAAI 2024 · similarity 0.90)
- Defense Against Adversarial Attacks on No-Reference Image Quality Models with Gradient Norm Regularization (CVPR 2024 · similarity 0.90)
- Backdoor Attacks Against No-Reference Image Quality Assessment Models via a Scalable Trigger (AAAI 2025 · similarity 0.88)
- Guardians of Image Quality: Benchmarking Defenses Against Adversarial Attacks on Image Quality Metrics (ICML 2025 · similarity 0.87)
- IOI: Invisible One-Iteration Adversarial Attack on No-Reference Image- and Video-Quality Metrics (ICML 2024 · similarity 0.87)
- GREAT Score: Global Robustness Evaluation of Adversarial Perturbation using Generative Models (NeurIPS 2024 · 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.
