Differential Privacy for Federated Learning
Abstract
This paper explores the application of differential privacy to federated learning, a distributed machine learning paradigm gaining prominence for its potential to leverage decentralized data while preserving privacy. Federated learning enables training of machine learning models across numerous devices or servers holding local data samples, without explicitly exchanging the data themselves. However, this decentralized approach introduces new privacy risks. This work proposes a method to integrate differential privacy into the federated learning process, specifically by adding noise to the model updates exchanged between participants. We demonstrate that this technique effectively limits the influence of any single participant's data on the global model, providing a quantifiable guarantee of privacy. The core contribution lies in the adaptation of differential privacy mechanisms to the unique challenges of federated learning, offering a robust solution for training machine learning models with sensitive data. The paper details the mathematical formulation of the proposed approach, including the privacy loss calculation and the impact of noise addition on model accuracy. We analyze the trade-off between privacy and accuracy, highlighting strategies for optimizing this balance. ---
Read the paper
Related studies
- Active Learning during Federated Learning for Object Detection
- Fairness-aware federated learning
- Decentralized Federated Learning
- Building a Personally Identifiable Information Recognizer in a Privacy Preserved Manner Using Automated Annotation and Federated Learning
- Preservation of Patient Level Privacy: Federated Classification and Calibration Models
- Efficient explainable federated learning
Go deeper in Evidence Hub
Evidence Hub adds a study data sheet (design, sample size, evidence level, key numbers), an AI summary and a chat grounded in the paper's text.