iDLG: Improved Deep Leakage from Gradients
arXiv:2001.02610
Abstract
It is widely believed that sharing gradients will not leak private training data in distributed learning systems such as Collaborative Learning and Federated Learning, etc. Recently, Zhu et al. presented an approach which shows the possibility to obtain private training data from the publicly shared gradients. In their Deep Leakage from Gradient (DLG) method, they synthesize the dummy data and corresponding labels with the supervision of shared gradients. However, DLG has difficulty in convergence and discovering the ground-truth labels consistently. In this paper, we find that sharing gradients definitely leaks the ground-truth labels. We propose a simple but reliable approach to extract accurate data from the gradients. Particularly, our approach can certainly extract the ground-truth labels as opposed to DLG, hence we name it Improved DLG (iDLG). Our approach is valid for any differentiable model trained with cross-entropy loss over one-hot labels. We mathematically illustrate how our method can extract ground-truth labels from the gradients and empirically demonstrate the advantages over DLG.
References in corpus (1)
Cited by in corpus (51)
- From Distributed Machine Learning to Federated Learning: A Survey
- Survey on Federated Learning Threats: concepts, taxonomy on attacks and defences, experimental study and challenges
- Threats to Federated Learning: A Survey
- Inverting Gradients -- How easy is it to break privacy in federated learning?
- HeteroFL: Computation and Communication Efficient Federated Learning for Heterogeneous Clients
- FedPD: A Federated Learning Framework with Optimal Rates and Adaptivity to Non-IID Data
- Federated Continual Learning via Knowledge Fusion: A Survey
- GRNN: Generative Regression Neural Network -- A Data Leakage Attack for Federated Learning
- Dataset Condensation with Gradient Matching
- A Framework for Evaluating Gradient Leakage Attacks in Federated Learning
- Privacy-Preserving Machine Learning: Methods, Challenges and Directions
- FedCG: Leverage Conditional GAN for Protecting Privacy and Maintaining Competitive Performance in Federated Learning
- LotteryFL: Personalized and Communication-Efficient Federated Learning with Lottery Ticket Hypothesis on Non-IID Datasets
- Joint Privacy Enhancement and Quantization in Federated Learning
- Label Leakage and Protection in Two-party Split Learning
- A Survey on Vulnerability of Federated Learning: A Learning Algorithm Perspective
- Enhanced Security and Privacy via Fragmented Federated Learning
- FedDef: Defense Against Gradient Leakage in Federated Learning-based Network Intrusion Detection Systems
- Get your Foes Fooled: Proximal Gradient Split Learning for Defense against Model Inversion Attacks on IoMT data
- Similarity-based Label Inference Attack against Training and Inference of Split Learning
- See through Gradients: Image Batch Recovery via GradInversion
- A Privacy Preserving System for Movie Recommendations Using Federated Learning
- Towards General Deep Leakage in Federated Learning
- Understanding Clipping for Federated Learning: Convergence and Client-Level Differential Privacy
- Privacy-preserving quantum federated learning via gradient hiding
- Federated Split Vision Transformer for COVID-19 CXR Diagnosis using Task-Agnostic Training
- Differentially Private Federated Learning for Cancer Prediction
- CAFE: Catastrophic Data Leakage in Vertical Federated Learning
- FedBoosting: Federated Learning with Gradient Protected Boosting for Text Recognition
- A Novel Attribute Reconstruction Attack in Federated Learning
- Layer-wise Characterization of Latent Information Leakage in Federated Learning
- On the Practicality of Differential Privacy in Federated Learning by Tuning Iteration Times
- Revealing and Protecting Labels in Distributed Training
- Federated Learning on Non-IID Data: A Survey
- Dynamic Differential-Privacy Preserving SGD
- TextHide: Tackling Data Privacy in Language Understanding Tasks
- PPCA: Privacy-preserving Principal Component Analysis Using Secure Multiparty Computation(MPC)
- Communication-Efficient Federated Learning with Compensated Overlap-FedAvg
- R-GAP: Recursive Gradient Attack on Privacy
- Privacy-preserving Collaborative Learning with Automatic Transformation Search
- Gradient Inversion with Generative Image Prior
- Understanding Training-Data Leakage from Gradients in Neural Networks for Image Classification
- Constrained Differentially Private Federated Learning for Low-bandwidth Devices
- Separation of Powers in Federated Learning
- Gain without Pain: Offsetting DP-injected Nosies Stealthily in Cross-device Federated Learning
- Optimizing the Numbers of Queries and Replies in Federated Learning with Differential Privacy
- SSGD: A safe and efficient method of gradient descent
- Dropout against Deep Leakage from Gradients
- Learning to Prevent Leakage: Privacy-Preserving Inference in the Mobile Cloud
- Efficient Byzantine-Resilient Stochastic Gradient Desce
- Privacy-Aware Document Visual Question Answering