Research map / Adversarial robustness and security / Attacks on text and graphs
Ask, Acquire, and Attack: Data-free UAP Generation using Class Impressions
Konda Reddy Mopuri, Phani Krishna Uppala, R. Venkatesh Babu
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 the data-free FFF objective with class impressions plus a learned generator | UAP generation (mean over 6 nets) · ILSVRC val | mean success rate: 56.58 -> 76.41 | Improved |
| use class impressions instead of real data | data-free UAP generator · ILSVRC val | mean success rate: 76.41 vs 84.67 with data | Mixed |
| add the learned generator on ResNet-152 target | FFF baseline · ILSVRC val ResNet-152 | mean success rate: 24.60 -> 53.73 | Improved |
| add diversity loss L_d | UAP generator · ILSVRC val VGG-F | labels covering 95% of predictions: 212 vs 173 | Improved |
| train the generator on an ensemble of three target CNNs | generator · black-box transfer to 3 unseen nets | MBBSR: 68.52 vs 60.34, 61.46, 52.43 | Improved |
| adversarially train target CNN with 50% adversarial mixture | GoogLeNet vs generator · ILSVRC val | success rate: 75.28 -> 62.51 | No clear effect |
| retrain generator on the adversarially finetuned network | generator · ILSVRC val | success rate: 68.72 | Hurt |
| change latent dimension from 10 to 50 or 100 (ablation) | generator · ILSVRC val | success rate | No clear effect |
| interpolate between latent points (ablation) | generator ResNet-152 · ILSVRC val | success rate | No clear effect |
Closest papers by meaning
- Data-free Universal Adversarial Perturbation with Pseudo-semantic Prior (CVPR 2025 · similarity 0.90)
- Data-Free Universal Attack by Exploiting the Intrinsic Vulnerability of Deep Models (AAAI 2025 · similarity 0.89)
- Improving Generalization of Universal Adversarial Perturbation via Dynamic Maximin Optimization (AAAI 2025 · similarity 0.88)
- Generate Universal Adversarial Perturbations for Few-Shot Learning (NeurIPS 2024 · similarity 0.85)
- Adversarial Initialization with Universal Adversarial Perturbation: A New Approach to Fast Adversarial Training (AAAI 2024 · similarity 0.85)
- Impartial Adversarial Distillation: Addressing Biased Data-Free Knowledge Distillation via Adaptive Constrained Optimization (AAAI 2024 · similarity 0.84)
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.
