Research map / Adversarial robustness and security / Attacks on text and graphs
CREDIT: Certified Ownership Verification of Deep Neural Networks Against Model Extraction Attacks
Bolin Shen, Zhan Cheng, Neil Zhenqiang Gong, Fan Yao, Yushun Dong
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 CREDIT Gaussian-noise defense instead of watermarking baselines | ResNet on CIFAR-10 · CIFAR-10 utility | accuracy: Backdooring 78.10 -> CREDIT 94.67 (vanilla 95.70) | Improved |
| use mutual-information verification (CREDIT) instead of fingerprint/watermark baselines | ResNet on CIFAR-10 · CIFAR-10 ownership verification | AUC: 58.64 -> 100.0 (best baseline Backdooring) | Improved |
| add CREDIT noise on graph backbones | GCN/GAT/SSGC on PROTEINS · PROTEINS utility | accuracy: 71.00 -> 72.50 (GCN) | Improved |
| add CREDIT noise on graph backbones | GCN on ENZYMES · ENZYMES utility | accuracy: 42.78 -> 40.61 | Hurt |
| replace baseline verification with CREDIT mutual-information estimation | VGG-16 verification stage · Verification time | seconds: 73.40 -> 22.23 (EWE) | Improved |
| increase backbone size from VGG-16 to ResNet-50 (ablation) | baselines vs CREDIT · Verification time | seconds: UAP 72.74 -> 180.3; CREDIT 22.23 -> 22.62 | Mixed |
| increase Gaussian noise sigma (ablation) | CREDIT · Certification error probabilities | gamma1 / gamma2 | Mixed |
| choose sigma by minimizing weighted utility and verification entropy (ablation) | CREDIT · sigma selection | objective: optimum at sigma = 0.27 | Improved |
Closest papers by meaning
- Data-Free Hard-Label Robustness Stealing Attack (AAAI 2024 · similarity 0.85)
- ZeroMark: Towards Dataset Ownership Verification without Disclosing Watermark (NeurIPS 2024 · similarity 0.85)
- SAME: Sample Reconstruction against Model Extraction Attacks (AAAI 2024 · similarity 0.84)
- Data Taggants: Dataset Ownership Verification Via Harmless Targeted Data Poisoning (ICLR 2025 · similarity 0.83)
- Dataset Ownership Verification in Contrastive Pre-trained Models (ICLR 2025 · similarity 0.83)
- Certified but Private: Scalable Zero-Knowledge Proofs for the Formal Verification of Neural Networks (NeurIPS 2026 · similarity 0.83)
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.
