Research map / Adversarial robustness and security / Black-box and transfer attacks
Open-Set Adversarial Defense with Clean-Adversarial Mutual Learning
Rui Shao, Pramuditha Perera, Pong Chi Yuen, Vishal M. Patel
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 |
|---|---|---|---|
| add CAML peer learner to DADL (ablation) | Ours · CIFAR-10 ablation (white-box PGD) | AUC-ROC: 71.69 -> 73.72 | Improved |
| add CAML peer learner to DADL (ablation) | Ours · CIFAR-10 ablation (white-box PGD) | closed-set accuracy: 73.13 -> 74.14 | Improved |
| add DADL to encoder (ablation) | Ours · CIFAR-10 ablation (white-box PGD) | AUC-ROC: 66.10 -> 69.86 | Improved |
| add DADL with decoder (ablation) | Ours · CIFAR-10 ablation (white-box PGD) | AUC-ROC: 67.34 -> 70.52 | Improved |
| add decoder to encoder alone (ablation) | Ours · CIFAR-10 ablation (white-box PGD) | closed-set accuracy: 69.90 -> 68.85 | Hurt |
| add self-supervised denoising SSD to Enc+DADL (ablation) | Ours · CIFAR-10 ablation (white-box PGD) | closed-set accuracy: 70.98 -> 73.13 | Improved |
| add CAML to Enc+Dec+SSD without DADL (ablation) | Ours · CIFAR-10 ablation (white-box PGD) | AUC-ROC: 67.04 -> 69.87 | Improved |
| use CAML instead of mixing clean and adversarial images (ablation) | Ours · CIFAR-10 white-box PGD | closed-set accuracy: 39.24 -> 76.25 | Improved |
| use CAML instead of mixing clean and adversarial images (ablation) | Ours · CIFAR-10 white-box PGD | open-set AUC-ROC: 56.10 -> 74.42 | Improved |
| add CAML on top of DADL | Ours · SVHN white-box PGD | closed-set accuracy: 80.01 -> 81.10 | Improved |
| add CAML on top of DADL | Ours · TinyImageNet white-box PGD | closed-set accuracy: 41.53 -> 43.77 | Improved |
| add CAML on top of DADL | Ours · CIFAR-10 black-box FGSM | open-set AUC-ROC: 75.35 -> 74.87 | Hurt |
Closest papers by meaning
- Open-set Adversarial Defense (ECCV 2020 · similarity 0.97)
- Adversarial Reciprocal Points Learning for Open Set Recognition (TPAMI 2021 · similarity 0.86)
- Learning Network Architecture for Open-Set Recognition (AAAI 2022 · similarity 0.86)
- OpenGAN: Open-Set Recognition via Open Data Generation (ICCV 2021 · similarity 0.85)
- Beyond Pretrained Features: Noisy Image Modeling Provides Adversarial Defense (NeurIPS 2023 · similarity 0.85)
- Adversarial Training Can Provably Improve Robustness: Theoretical Analysis of Feature Learning Process Under Structured Data (ICLR 2025 · 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.
