Research map / Adversarial robustness and security / Attacks on text and graphs
Art of Singular Vectors and Universal Adversarial Perturbations
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 the Jacobian singular-vector method to construct a universal perturbation from 64 images | VGG-19 block2_pool · ILSVRC 2012 validation | fooling rate: 0.60 | Helped |
| Increase the batch size used to construct the perturbation | SingularFool on VGG-19 block2_pool · ILSVRC 2012 validation | fooling rate | No clear effect |
| Construct the perturbation using as few as 16 images | SingularFool on VGG-19 block2_pool · ILSVRC 2012 validation | fooling rate: 56 % | Helped |
| Use larger q values | SingularFool on VGG-19 block2_pool · ILSVRC 2012 validation | fooling rate | Helped |
| Set q approximately to 5 rather than using q=10 | SingularFool on VGG-19 block2_pool · ILSVRC 2012 validation | fooling rate | Helped |
| Use a perturbation computed on VGG-16 and test it on VGG-19 | SingularFool · ILSVRC 2012 validation, VGG-19 | fooling rate: 0.60 | Helped |
| Use a perturbation computed on VGG-19 and test it on VGG-16 | SingularFool · ILSVRC 2012 validation, VGG-16 | fooling rate: 0.48 | Mixed |
| Use a perturbation computed on ResNet50 and test it on VGG-19 | SingularFool · ILSVRC 2012 validation, VGG-19 | fooling rate: 0.47 | Helped |
| Use SingularFool rather than the UAP baseline at the reported comparison settings | VGG-19 block2_pool · ILSVRC 2012 validation | fooling rate: roughly 60% ; roughly 20 % | Helped |
| Use SingularFool rather than UAP as the number of construction images increases | VGG-19 block2_pool · ILSVRC 2012 validation | fooling rate: 63% | Helped |
Closest papers by meaning
- Contrasting Adversarial Perturbations: The Space of Harmless Perturbations (AAAI 2025 · similarity 0.89)
- Theoretical Understanding of Learning from Adversarial Perturbations (ICLR 2024 · similarity 0.88)
- Data-Free Universal Attack by Exploiting the Intrinsic Vulnerability of Deep Models (AAAI 2025 · similarity 0.88)
- Improving Generalization of Universal Adversarial Perturbation via Dynamic Maximin Optimization (AAAI 2025 · similarity 0.87)
- Wide Two-Layer Networks can Learn from Adversarial Perturbations (NeurIPS 2024 · similarity 0.87)
- Adversarial Perturbations Are Formed by Iteratively Learning Linear Combinations of the Right Singular Vectors of the Adversarial Jacobian (ICML 2025 · similarity 0.87)
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.