Research map / Adversarial robustness and security / Attacks on text and graphs
A Self-supervised Approach for Adversarial Robustness
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 |
|---|---|---|---|
| Add NRP input purification | IncRes-v2 ensemble · ImageNet black-box ensemble DIM_TI attack | attack success rate: 82.6% to 31.9% | Helped |
| Use NRP as a plug-in purifier | Classification, detection, and segmentation systems · CDA across ImageNet, Comics, and Paintings | accuracy | Helped |
| Add feature loss to the purifier objective (ablation) | NRP · ImageNet-NeurIPS black-box ensemble attack | top-1 accuracy | Helped |
| Add pixel loss to the purifier objective (ablation) | NRP · ImageNet-NeurIPS black-box ensemble attack | top-1 accuracy | Helped |
| Use relativistic average GAN loss instead of vanilla GAN loss (ablation) | NRP · ImageNet-NeurIPS black-box ensemble attack | top-1 accuracy | Helped |
| Train the purifier as a Gaussian denoiser (GNP) instead of using SSP adversarial training (ablation) | NRP · Translation-invariant attacks | robustness | Hurt |
| Train the purifier on FGSM adversaries (FGSP) instead of SSP adversaries (ablation) | NRP · Adversarial defense | robustness | Helped |
| Generate training perturbations in VGG perceptual feature space rather than other tested network feature spaces (ablation) | SSP and NRP · ImageNet-NeurIPS transferability | attack transferability | Helped |
| Distort VGG conv3.3 features rather than other tested layers (ablation) | SSP · ImageNet-NeurIPS transferability | attack transferability | Helped |
| Increase representation loss at multiple network layers rather than a single layer (ablation) | SSP · ImageNet-NeurIPS | attack success rate | No clear effect |
| Remove the purifier's input-to-output skip connection (ablation) | NRP generator · Adversarial purification | robustness | Helped |
Closest papers by meaning
- A Direct Approach to Robust Deep Learning Using Adversarial Networks (ICLR 2019 · similarity 0.90)
- Benchmarking Adversarial Robustness on Image Classification (CVPR 2020 · similarity 0.89)
- Towards Deep Learning Models Resistant to Adversarial Attacks (ICLR 2018 · similarity 0.89)
- Improving Adversarial Robustness Requires Revisiting Misclassified Examples (ICLR 2020 · similarity 0.89)
- Adversarial Training Can Provably Improve Robustness: Theoretical Analysis of Feature Learning Process Under Structured Data (ICLR 2025 · similarity 0.88)
- Enhancing Adversarial Transferability with Checkpoints of a Single Model's Training (CVPR 2025 · similarity 0.88)
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.