Research map / Adversarial robustness and security / Certified and empirical robustness
Provably Adversarially Robust Detection of Out-of-Distribution Data (Almost) for Free
Alexander Meinke, Julian Bitterwolf, Matthias Hein
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 certified binary discriminator to OE classifier (ProoD vs OE) | OE ResNet18 · CIFAR10 vs SVHN | AAUC: 0.0 -> 54.1 | Improved |
| add certified binary discriminator to OE classifier (ProoD vs OE) | OE ResNet18 · CIFAR10 test | accuracy: 94.91 -> 94.99 | No clear effect |
| ProoD vs GOOD80 (architecture-unconstrained classifier) | GOOD80 · CIFAR10 test | accuracy: 87.39 -> 94.99 | Improved |
| ProoD vs GOOD80 on guaranteed AUC | GOOD80 · CIFAR10 vs CIFAR100 | GAUC: 47.1 -> 46.1 | Hurt |
| ProoD vs GOOD80 on guaranteed AUC | GOOD80 · CIFAR10 vs SVHN | GAUC: 43.4 -> 53.3 | Improved |
| ProoD vs adversarially trained ATOM | ATOM · CIFAR10 vs SVHN | AAUC: 24.1 -> 54.1 | Improved |
| ProoD vs ATOM on CIFAR100 | ATOM · CIFAR100 vs CIFAR10 | AAUC: 50.3 -> 17.4 | Hurt |
| ProoD vs ATOM on accuracy CIFAR100 | ATOM · CIFAR100 test | accuracy: 68.32 -> 77.16 | Improved |
| add classifier to discriminator (ProoD-Disc to ProoD) for clean OOD detection (ablation) | ProoD-Disc · CIFAR10 vs CIFAR100 | AUC: 62.9 -> 89.8 | Improved |
| semi-joint training of classifier through the fixed discriminator instead of separate training (ablation) | ProoD · CIFAR10 | guaranteed AUC | Improved |
| positive bias shift Delta on discriminator output (sweep 0 to 6) (ablation) | ProoD · CIFAR10/CIFAR100/R.ImgNet | AUC vs guaranteed AUC trade-off | Mixed |
| enforce negative output-layer weights of discriminator (ablation) | ProoD · CIFAR10 test | accuracy and OOD AUC | No clear effect |
Closest papers by meaning
- Certifiably Adversarially Robust Detection of Out-of-Distribution Data (NeurIPS 2020 · similarity 0.92)
- Adversarially Robust Out-of-Distribution Detection Using Lyapunov-Stabilized Embeddings (ICLR 2025 · similarity 0.91)
- Your Out-of-Distribution Detection Method is Not Robust! (NeurIPS 2022 · similarity 0.90)
- OODRobustBench: a Benchmark and Large-Scale Analysis of Adversarial Robustness under Distribution Shift (ICML 2024 · similarity 0.87)
- Leveraging Perturbation Robustness to Enhance Out-of-Distribution Detection (CVPR 2025 · similarity 0.87)
- Splitting & Integrating: Out-of-Distribution Detection via Adversarial Gradient Attribution (ICML 2025 · similarity 0.86)
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.
