Research map / Adversarial robustness and security / Black-box and transfer attacks
Learning to Attack: Adversarial Transformation Networks
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 |
|---|---|---|---|
| decrease reconstruction weight beta from 0.010 to 0.001 (ablation) | ATNa · MNIST | attack success rate: 69.1% -> 95.9% | Helped |
| train ATN against 3 target networks instead of 1 (ablation) | ATNb · MNIST transfer to unseen Ca3 | rank-1 success: 29% -> 64% | Helped |
| AAE (autoencoding) ATN instead of perturbation ATN (ablation) | IR2-Base-Deconv ATN · ImageNet IR2 | adversarial top-1 success: 56.5% -> 92.1% | Helped |
| choice of ATN architecture (FC vs conv vs deconv) (ablation) | ATN · MNIST | attack success rate: 84.1% vs 77.7% vs 82.5% | Mixed |
Closest papers by meaning
- Countering Adversarial Images using Input Transformations (ICLR 2018 · similarity 0.90)
- Detecting Adversarial Examples Through Image Transformation (AAAI 2018 · similarity 0.89)
- With Friends Like These, Who Needs Adversaries? (NeurIPS 2018 · similarity 0.88)
- Beyond ImageNet Attack: Towards Crafting Adversarial Examples for Black-box Domains (ICLR 2022 · similarity 0.88)
- Improving the Transferability of Adversarial Samples With Adversarial Transformations (CVPR 2021 · similarity 0.88)
- Enhancing Transformation-Based Defenses Against Adversarial Attacks with a Distribution Classifier (ICLR 2020 · 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.