Understanding Deep Gradient Leakage via Inversion Influence Functions
Preprint
Abstract
Deep Gradient Leakage (DGL) is a highly effective attack that recovers private training images from gradient vectors. This attack casts significant privacy challenges on distributed learning from clients with sensitive data, where clients are required to share gradients. Defending against such attacks requires but lacks an understanding of when and how privacy leakage happens, mostly because of the black-box nature of deep networks. In this paper, we propose a novel Inversion Influence Function (I$^2$F) that establishes a closed-form connection between the recovered images and the private gradients by implicitly solving the DGL problem. Compared to directly solving DGL, I$^2$F is scalable for analyzing deep networks, requiring only oracle access to gradients and Jacobian-vector products. We empirically demonstrate that I$^2$F effectively approximated the DGL generally on different model architectures, datasets, modalities, attack implementations, and perturbation-based defenses. With this novel tool, we provide insights into effective gradient perturbation directions, the unfairness of privacy protection, and privacy-preferred model initialization. Our codes are provided in https://github.com/illidanlab/inversion-influence-function.Keywords
All Related Versions
This publication has 11 references indexed in Scilit:
- More Than Enough is Too Much: Adaptive Defenses Against Gradient Leakage in Production Federated LearningIEEE/ACM Transactions on Networking, 2024
- Review on security of federated learning and its application in healthcareFuture Generation Computer Systems, 2023
- Federated Learning for Healthcare: Systematic Review and Architecture ProposalACM Transactions on Intelligent Systems and Technology, 2022
- Federated learning for predicting clinical outcomes in patients with COVID-19Nature Medicine, 2021
- Soteria: Provable Defense against Privacy Leakage in Federated Learning from Representation Perspective2021
- End-to-end privacy preserving deep learning on multi-institutional medical imagingNature Machine Intelligence, 2021
- TAG: Gradient Attack on Transformer-based Language Models2021
- Federated Learning for Healthcare InformaticsJournal of Healthcare Informatics Research, 2020
- Deep Residual Learning for Image Recognition2016
- Gradient-based learning applied to document recognitionProceedings of the IEEE, 1998