USENIX Security2025Top-tier venue
Task-Oriented Training Data Privacy Protection for Cloud-based Model Training
Zhiqiang Wang, Jiahui Hou, Haifeng Sun, Jingmiao Zhang, Yunhao Yao, Haikuo Yu, Xiang-Yang Li
Abstract
Cloud-based model training presents significant privacy challenges, as users must upload personal data for training highperformance models. Once uploaded, this data goes beyond the user's control and could be misused for other purposes. Users need tools to control the usage scope of the uploaded training data, preventing unauthorized training without compromising authorized training. Unfortunately, existing solutions overlook this issue.
In this paper, we propose and achieve a unique privacyutility goal tailored for cloud-based model training, considering both user demand and legal requirements. Our approach provides task-level control of training data usage, simultaneously ensuring each protected data exhibits noticeable visual changes to address fundamental privacy concerns. We introduce carefully designed noise to each training data for privacy protection. These noises are designed to provide visual protection while minimizing the shifts in the feature domain through adversarial optimization. By adjusting the correlation between noise and class labels, we guide the model to learn the correct features for the target task while preventing unauthorized privacy task training. Additionally, we introduce the overflow matrix for compatibility with existing encoding and transmission frameworks. Real-world experiments demonstrate that it can simultaneously protect visual privacy (SSIM is 0.028) and prevent unauthorized model training (protection success rate achieved 100%), while the accuracy of the target task model is slightly reduced by about 1.8%.
Ask Lune
Know this paper in minutes, not an afternoon.
Lune can read every page of it and the top-tier papers that cite it, and gives you its claims, its evidence and its limits, each one cited to the paper it comes from. No skimming, no AI slop.
Lineage
- Unlearnable Examples: Making Personal Data Unexploitable cites Membership Inference Attacks Against Machine Learning Models
- Unlearnable Examples: Making Personal Data Unexploitable cites Deep Learning with Differential Privacy
- Membership Inference Attacks Against Machine Learning Models cites Deep Learning with Differential Privacy
- Dataset Distillation by Matching Training Trajectories cites Dataset Condensation with Differentiable Siamese Augmentation
Builds on17
- Deep Learning with Differential PrivacyMartín Abadi, Andy Chu, Ian J. Goodfellow, H. Brendan McMahan et al.CCS 2016 · 7,620 citations
- Membership Inference Attacks Against Machine Learning ModelsReza Shokri, Marco Stronati, Congzheng Song, Vitaly ShmatikovS&P 2017 · 5,137 citations
- Dataset Condensation with Differentiable Siamese AugmentationBo Zhao, Hakan BilenICML 2021 · 390 citations
- Unlearnable Examples: Making Personal Data UnexploitableHanxun Huang, Xingjun Ma, Sarah Monazam Erfani, James Bailey et al.ICLR 2021 · 255 citations
- Dataset Distillation by Matching Training TrajectoriesGeorge Cazenavette, Tongzhou Wang, Antonio Torralba, Alexei A. Efros et al.CVPR 2022 · 198 citations
Related papers
- Privacy Adversarial Network: Representation Learning for Mobile Data PrivacySicong Liu, Junzhao Du, Anshumali Shrivastava, Lin ZhongUbiComp 2020 · 46 citations
- Towards Safe Machine Unlearning: A Paradigm that Mitigates Performance DegradationShanshan Ye, Jie Lu, Guangquan ZhangWWW 2025 · 13 citations
- You Can Use But Cannot Recognize: Preserving Visual Privacy in Deep Neural NetworksQiushi Li, Yan Zhang, Ju Ren, Qi Li et al.NDSS 2024
- Once-for-all: Efficient Visual Face Privacy Protection via Person-specific VeilsZixuan Yang, Yushu Zhang, Tao Wang, Zhongyun Hua et al.ACM MM 2024 · 2 citations
- Unlearnable 3D Point Clouds: Class-wise Transformation Is All You NeedXianlong Wang, Minghui Li, Wei Liu, Hangtao Zhang et al.NeurIPS 2024 · 23 citations
