Data-Centric AI Intellectual Property Protection

This repository introduces research topics related to protecting the intellectual property (IP) of AI from a data-centric perspective. Such topics include data-centric model IP protection, data authorization protection, data copyright protection, and any other data-level technologies that protect the IP of AI. More content is coming, and in the end, we care about your uniqueness!!!

Data-Centric Model Protection

Verify your ownership of a certain model via certain data and authorize the usage of your model to certain data.

Image Data

• Non-Transferable Learning: A New Approach for Model Ownership Verification and Applicability Authorization
  • [paper] [code]
  • ICLR 2022 Oral
• Model Barrier: A Compact Un-Transferable Isolation Domain for Model Intellectual Property Protection
  • [paper] [code]
  • CVPR 2023
• Domain Specified Optimization for Deployment Authorization
  • [paper]
  • ICCV 2023

Other Data

• Unsupervised Non-transferable Text Classification
  • [paper] [code]
  • EMNLP 2022

Data Authorization Protection (namely unlearnable data or examples)

Prevent unauthorized data usage of model training, usually achieved by decreasing the model performance via poisoning attacks.

Image Data

• Unlearnable Examples: Making Personal Data Unexploitable
  • [paper] [code]
  • ICLR 2021
• Going Grayscale: The Road to Understanding and Improving Unlearnable Examples
  • [paper] [code]
  • Arxiv 2021
• Robust Unlearnable Examples: Protecting Data Against Adversarial Learning
  • [paper] [code]
  • ICLR 2022
• Self-Ensemble Protection: Training Checkpoints Are Good Data Protectors
  • [paper] [code]
  • ICLR 2023
• Transferable Unlearnable Examples
  • [paper] [code]
  • ICLR 2023
• LAVA: Data Valuation without Pre-Specified Learning Algorithms
  • [paper] [code]
  • ICLR 2023 Spotlight
• Unlearnable Clusters: Towards Label-Agnostic Unlearnable Examples
  • [paper] [code]
  • CVPR 2023
• Universal Unlearnable Examples: Cluster-wise Perturbations without Label-consistency
  • [paper]
  • ICLR 2023 submission
• Unlearnable Examples Give a False Sense of Security: Piercing through Unexploitable Data with Learnable Examples
  • [paper]
  • Arxiv 2023
• Towards Generalizable Data Protection With Transferable Unlearnable Examples
  • [paper]
  • Arxiv 2023
• CUDA: Convolution-Based Unlearnable Datasets
  • [paper] [code]
  • CVPR 2023
• Raising the Cost of Malicious AI-Powered Image Editing
  • [paper] [code]
  • Arxiv 2023
• Learning the Unlearnable: Adversarial Augmentations Suppress Unlearnable Example Attacks
  • [paper] [code]
  • Arxiv 2023
• Towards Generalizable Data Protection With Transferable Unlearnable Examples
  • [paper]
  • Arxiv 2023
• The Devil's Advocate: Shattering the Illusion of Unexploitable Data using Diffusion Models
  • [paper]
  • Arxiv 2023
• GLAZE: Protecting Artists from Style Mimicry by Text-to-Image Models
  • [paper] [App]
  • Usenix Security 2023
• Flew Over Learning Trap: Learn Unlearnable Samples by Progressive Staged Training
  • [paper] [code]
  • Arxiv 2023
• Segue: Side-information Guided Generative Unlearnable Examples for Facial Privacy Protection in Real World
  • [paper]
  • Arxiv 2023
• What Can We Learn from Unlearnable Datasets?
  • [paper] [code]
  • NeurIPS 2023

Other Data

• Unlearnable Examples: Protecting Open-Source Software from Unauthorized Neural Code Learning
  • [paper] [code]
  • SEKE 2022
• WaveFuzz: A Clean-Label Poisoning Attack to Protect Your Voice
  • [paper]
  • Arxiv 2023
• Unlearnable Graph: Protecting Graphs from Unauthorized Exploitation
  • [paper]
  • Poster at NDSS 2023
• Securing Biomedical Images from Unauthorized Training with Anti-Learning Perturbation
  • [paper]
  • Poster at NDSS 2023
• UPTON: Unattributable Authorship Text via Data Poisoning
  • [paper]
  • Arxiv 2023
• GraphCloak: Safeguarding Task-specific Knowledge within Graph-structured Data from Unauthorized Exploitation
  • [paper]
  • Arxiv 2023
• Make Text Unlearnable: Exploiting Effective Patterns to Protect Personal Data
  • [paper]
  • Arxiv 2023

Data Copyright Protection

Verify your ownership of certain data via black-box model access.

Image Data

• Radioactive data: tracing through training
  • [paper] [code]
  • ICML 2020
• Tracing Data through Learning with Watermarking
  • [paper]
  • Proceedings of the 2021 ACM Workshop on Information Hiding and Multimedia Security
• On the Effectiveness of Dataset Watermarking
  • [paper]
  • Proceedings of the 2022 ACM on International Workshop on Security and Privacy Analytics
• Untargeted Backdoor Watermark: Towards Harmless and Stealthy Dataset Copyright Protection
  • [paper] [code]
  • NeurIPS 2022 Oral
• Did You Train on My Dataset? Towards Public Dataset Protection with Clean-Label Backdoor Watermarking
  • [paper]
  • Arxiv 2023
• On the Effectiveness of Dataset Watermarking in Adversarial Settings
  • [paper]
  • Proceedings of CODASPY-IWSPA 2022
• Anti-Neuron Watermarking: Protecting Personal Data Against Unauthorized Neural Networks
  • [paper]
  • ECCV 2022
• Data Isotopes for Data Provenance in DNNs
  • [paper] [code]
  • Arxiv 2022
• Watermarking for Data Provenance in Object Detection
  • [paper]
  • 2022 IEEE Applied Imagery Pattern Recognition Workshop (AIPR)
• Reclaiming the Digital Commons: A Public Data Trust for Training Data
  • [paper]
  • AIES 2023
• MedLocker: A Transferable Adversarial Watermarking for Preventing Unauthorized Analysis of Medical Image Dataset
  • [paper]
  • Arxiv 2023
• How to Detect Unauthorized Data Usages in Text-to-image Diffusion Models
  • [paper]
  • Arxiv 2023
• FT-Shield: A Watermark Against Unauthorized Fine-tuning in Text-to-Image Diffusion Models
  • [paper]
  • Arxiv 2023
• Domain Watermark: Effective and Harmless Dataset Copyright Protection is Closed at Hand
  • [paper] [code]
  • NeurIPS 2023
• DiffusionShield: A Watermark for Copyright Protection against Generative Diffusion Models
  • [paper]
  • Arxiv 2023

Other Data

• CoProtector: Protect Open-Source Code against Unauthorized Training Usage with Data Poisoning
  • [paper] [code]
  • WWW 2022
• CodeMark: Imperceptible Watermarking for Code Datasets against Neural Code Completion Models
  • [paper]
  • FSE 2023
• Watermarking Classification Dataset for Copyright Protection
  • [paper]
  • Arxiv 2023