Research map / Adversarial robustness and security / Attacks on text and graphs
Passport-aware Normalization for Deep Model Protection
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 |
|---|---|---|---|
| add independent passport-aware normalization branch (separate statistics) instead of shared (ablation) | passport model · CIFAR100 ResNet-18 | deployment accuracy: 2.05 -> 73.84 | Helped |
| add independent passport-aware normalization branch (ablation) | passport model · CIFAR10 ResNet-18 | deployment accuracy: 17.46 -> 94.04 | Helped |
| make gamma1, beta1 learnable (config C vs B) (ablation) | passport model · CIFAR100 ResNet-18 | deployment accuracy: 73.84 -> 74.40 | Helped |
| make gamma1, beta1 learnable (config C vs B) (ablation) | passport model · CIFAR10 AlexNet | deployment accuracy: 89.54 -> 90.00 | Helped |
| use passport-aware normalization with BN instead of Fan et al. method (BN) | ResNet-18 · CIFAR100 | deployment accuracy: 1.90 -> 74.40 | Helped |
| apply passport-aware normalization only to last three norm layers (ablation) | ResNet-18 · CIFAR100 | deployment accuracy: 74.40 -> 76.22 | Helped |
| reduce passport branch activation ratio from 50% to 10% (ablation) | PointNet · ShapeNet | deployment accuracy: 99.31 -> 99.36 | No clear effect |
| replace BN with GN in target model (required by baseline) | ResNet-18 · CIFAR100 | accuracy: 76.60 -> 72.43 | Hurt |
| use passport-aware normalization vs baseline against ambiguity attack II (10% flipped) | ResNet-18 · CIFAR100 | attacker accuracy: 55.73 -> 6.21 | Helped |
| train passport branches alternately vs simultaneously (ablation) | passport-aware normalization · image and point tasks | accuracy | No clear effect |
Closest papers by meaning
- Steganographic Passport: An Owner and User Verifiable Credential for Deep Model IP Protection Without Retraining (CVPR 2024 · similarity 0.85)
- MAP: MAsk-Pruning for Source-Free Model Intellectual Property Protection (CVPR 2024 · similarity 0.84)
- IPRemover: A Generative Model Inversion Attack against Deep Neural Network Fingerprinting and Watermarking (AAAI 2024 · similarity 0.83)
- SDBF: Steep-Decision-Boundary Fingerprinting for Hard-Label Tampering Detection of DNN Models (CVPR 2025 · similarity 0.82)
- Graph-Embedded Structure-Aware Perceptual Hashing for Neural Network Protection and Piracy Detection (CVPR 2025 · similarity 0.81)
- Intersecting-Boundary-Sensitive Fingerprinting for Tampering Detection of DNN Models (ICML 2024 · similarity 0.81)
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.