Research map / Adversarial robustness and security / Certified and empirical robustness
A Simple Unified Framework for Detecting Out-of-Distribution Samples and Adversarial Attacks
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 |
|---|---|---|---|
| Replace the maximum softmax posterior confidence with a Mahalanobis distance score | ResNet trained on CIFAR-10 · SVHN OOD | TNR at TPR 95%: 86.6 / 96.4 | Helped |
| Use Mahalanobis scores instead of LID scores | ResNet trained on CIFAR-10 · CW detection | AUROC: 82.2% to 95.8% | Helped |
| Add input pre-processing to the Mahalanobis score (ablation) | Mahalanobis detector · SVHN OOD (CIFAR-10 ResNet) | TNR at TPR 95%: 54.51 -> 92.26 | Helped |
| Add feature ensemble to the Mahalanobis score (ablation) | Mahalanobis detector · SVHN OOD (CIFAR-10 ResNet) | TNR at TPR 95%: 54.51 -> 91.45 | Helped |
| Combine feature ensemble with input pre-processing (ablation) | Mahalanobis detector · SVHN OOD (CIFAR-10 ResNet) | TNR at TPR 95%: 54.51 -> 96.42 | Helped |
| Tune on FGSM samples and use the detector on other attacks | Mahalanobis detector · BIM, DeepFool, and CW detection | AUROC | Mixed |
Closest papers by meaning
- Training Confidence-calibrated Classifiers for Detecting Out-of-Distribution Samples (ICLR 2018 · similarity 0.89)
- Extreme Classification via Adversarial Softmax Approximation (ICLR 2020 · similarity 0.88)
- Towards In-Distribution Compatible Out-of-Distribution Detection (AAAI 2023 · similarity 0.87)
- CASE: Exploiting Intra-class Compactness and Inter-class Separability of Feature Embeddings for Out-of-Distribution Detection (AAAI 2024 · similarity 0.86)
- Splitting & Integrating: Out-of-Distribution Detection via Adversarial Gradient Attribution (ICML 2025 · similarity 0.86)
- DOS: Diverse Outlier Sampling for Out-of-Distribution Detection (ICLR 2024 · 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.