Research map / Adversarial robustness and security / Attacks on text and graphs
Non-Transferable Learning: A New Approach for Model Ownership Verification and Applicability Authorization
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 |
|---|---|---|---|
| Apply Target-Specified NTL rather than supervised learning | VGG-11 trained on MNIST, evaluated on USPS · MNIST→USPS (MT→US) | accuracy (%): 86.4 -> 14.5 | Hurt |
| Apply Target-Specified NTL rather than supervised learning | VGG-11 trained on MNIST · MNIST source domain | accuracy (%): 98.9 -> 97.9 | No clear effect |
| Apply Source-Only NTL rather than supervised learning | VGG-11 trained on MNIST, evaluated on USPS · MNIST→USPS (MT→US) | accuracy (%): 86.4 -> 13.8 | Hurt |
| Apply Source-Only NTL rather than supervised learning | VGG-11 trained on MNIST · MNIST source domain | accuracy (%): 98.9 -> 98.9 | No clear effect |
| Use the trigger-patch ownership verification behavior | Target-Specified NTL trained on MNIST · MNIST, test with versus without patch | accuracy (%): 10.2/98.9 | Mixed |
| Apply watermark removal methods to the NTL ownership behavior | Target-Specified NTL trained on MNIST · MNIST, test with patch | accuracy (%): 10.2 -> 9.9; 10.2 -> 10.4; 10.2 -> 10.9; 10.2 -> 10.8; 10.2 -> 11.0; 10.2 -> 10.0 | No clear effect |
| Authorize only the patched MNIST source domain | Source-Only NTL applicability authorization · MNIST authorized domain versus other domains | accuracy (%): 98.5; 13.0 | Mixed |
| Use Source-Only NTL rather than supervised learning | VGG-11 trained on MNIST, evaluated on MNIST-M · MNIST→MNIST-M (MT→MM) | accuracy (%): 57.4 -> 13.4 | Hurt |
| Use Target-Specified NTL rather than supervised learning | VGG-11 trained on MNIST-M, evaluated on MNIST · MNIST-M→MNIST (MM→MT) | accuracy (%): 97.0 -> 11.7 | Hurt |
Closest papers by meaning
- Catch-Only-One: Non-Transferable Examples for Model-Specific Authorization (NeurIPS 2026 · similarity 0.87)
- ZeroMark: Towards Dataset Ownership Verification without Disclosing Watermark (NeurIPS 2024 · similarity 0.87)
- MFL-Owner: Ownership Protection for Multi-modal Federated Learning via Orthogonal Transform Watermark (AAAI 2025 · similarity 0.86)
- Your Transferability Barrier is Fragile: Free-Lunch for Transferring the Non-Transferable Learning (CVPR 2024 · similarity 0.86)
- Jailbreaking the Non-Transferable Barrier via Test-Time Data Disguising (CVPR 2025 · similarity 0.86)
- Defense against Model Extraction Attack by Bayesian Active Watermarking (ICML 2024 · similarity 0.85)
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.