Research map / Label-efficient and robust learning / Domain adaptation and generalization
Say No to Freeloader: Protecting Intellectual Property of Your Deep Model
Lianyu Wang, Meng Wang, Huazhu Fu, Daoqiang Zhang
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 L_t loss to the CUPI-Domain generators (ablation) | SL + CUPI generators · MNIST -> SVHN | unauthorized domain accuracy: 11.4 -> 9.4 | Improved |
| add L_i loss to the CUPI-Domain generators (ablation) | SL + CUPI generators · MNIST -> SVHN | unauthorized domain accuracy: 9.4 -> 9.4 | No clear effect |
| add L_i loss to the CUPI-Domain generators (ablation) | SL + CUPI generators · CIFAR10 -> STL10 | unauthorized domain accuracy: 11.4 -> 11.2 | Improved |
| add DIMB with L_Stl and L_Dis (ablation) | CUPI-Domain generators · MNIST -> SVHN | unauthorized domain accuracy: 9.4 -> 6.1 | Improved |
| add L_Stl loss (ablation) | CUPI-Domain with DIMB · MNIST -> SVHN | unauthorized domain accuracy: 6.3 -> 6.1 | Improved |
| add L_Dis loss (ablation) | CUPI-Domain with DIMB · VisDA-T -> VisDA-V | unauthorized domain accuracy: 9.6 -> 7.5 | Improved |
| add CUPI-Domain generators to supervised learning (ablation) | SL · MNIST -> SVHN | unauthorized domain accuracy: 38.2 -> 9.4 | Improved |
| use CUPI-Domain instead of CUTI-Domain, target-specified | digit datasets · Digit datasets | authorized domain accuracy drop: 0.13 -> 0.02 | Improved |
| use CUPI-Domain instead of NTL, target-specified | digit datasets · Digit datasets | unauthorized domain accuracy drop: 46.48 -> 56.83 | Improved |
| use CUPI-Domain instead of CUTI-Domain, target-free | digit datasets · Digit datasets | unauthorized domain accuracy drop: 53.47 -> 55.10 | Improved |
| use CUPI-Domain instead of NTL in applicability authorization scheme 1 | digits, CIFAR10&STL10, VisDA · Scheme-1 | accuracy drop: 81.03 -> 88.57 | Improved |
| use CUPI-Domain instead of CUTI in applicability authorization scheme 2 | digits, CIFAR10&STL10, VisDA · Scheme-2 | accuracy drop: 85.85 -> 85.50 | No clear effect |
Closest papers by meaning
- MAP: MAsk-Pruning for Source-Free Model Intellectual Property Protection (CVPR 2024 · similarity 0.86)
- Vision-Language Model IP Protection via Prompt-based Learning (CVPR 2025 · similarity 0.86)
- EncryIP: A Practical Encryption-Based Framework for Model Intellectual Property Protection (AAAI 2024 · similarity 0.83)
- Graph-Embedded Structure-Aware Perceptual Hashing for Neural Network Protection and Piracy Detection (CVPR 2025 · similarity 0.82)
- Disentangled Style Domain for Implicit $z$-Watermark Towards Copyright Protection (NeurIPS 2024 · similarity 0.82)
- Steganographic Passport: An Owner and User Verifiable Credential for Deep Model IP Protection Without Retraining (CVPR 2024 · similarity 0.82)
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.
