Research map / Adversarial robustness and security / Attacks on text and graphs
SparseFool: A Few Pixels Make a Big Difference
Check 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 SparseFool instead of JSMA | LeNet · MNIST test | perturbation (%): 4.85 -> 1.66 | Helped |
| Use SparseFool instead of JSMA | LeNet · MNIST test | time (sec): 0.66 -> 0.14 | Helped |
| Use SparseFool instead of One-pixel attack | LeNet · MNIST test | time (sec): 310.2 -> 0.14 | Helped |
| Use SparseFool instead of JSMA | VGG-19 · CIFAR-10 test | perturbation (%): 2.25 -> 1.07 | Helped |
| Use SparseFool instead of JSMA | VGG-19 · CIFAR-10 test | time (sec): 6.28 -> 0.34 | Helped |
| Use SparseFool instead of One-pixel attack | VGG-19 · CIFAR-10 test | perturbation (%): 0.15 -> 1.07 | Hurt |
| Use SparseFool instead of One-pixel attack | VGG-19 · CIFAR-10 test | time (sec): 102.7 -> 0.34 | Helped |
| Constrain perturbed pixels to a ±δ range around their original values (ablation) | SparseFool · ImageNet | perceptibility | Helped |
| Increase λ from values close to 1 (ablation) | SparseFool · MNIST, CIFAR-10, and ImageNet | fooling rate | Helped |
| Increase λ from values close to 1 (ablation) | SparseFool · MNIST, CIFAR-10, and ImageNet | sparsity | Hurt |
| Train with ℓ∞ adversarial perturbations (ablation) | ResNet-18 · CIFAR-10 | perturbation (%) under SparseFool attack: 1.27% -> 2.44% | Hurt |
| Train with ℓ∞ adversarial perturbations (ablation) | ResNet-18 · CIFAR-10 | robustness to sparse perturbations | No clear effect |
Closest papers by meaning
- Transferable Structural Sparse Adversarial Attack Via Exact Group Sparsity Training (CVPR 2024 · similarity 0.89)
- GSE: Group-wise Sparse and Explainable Adversarial Attacks (ICLR 2025 · similarity 0.88)
- BRUSLEATTACK: A QUERY-EFFICIENT SCORE- BASED BLACK-BOX SPARSE ADVERSARIAL ATTACK (ICLR 2024 · similarity 0.87)
- Sparse Cocktail: Every Sparse Pattern Every Sparse Ratio All At Once (ICML 2024 · similarity 0.86)
- SuperDeepFool: a new fast and accurate minimal adversarial attack (NeurIPS 2024 · similarity 0.85)
- Towards Efficient Training and Evaluation of Robust Models against $l_0$ Bounded Adversarial Perturbations (ICML 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.