Survey on Privacy-Preserving Techniques for Federated Learning
Abstract
: Federated learning (FL) is a process that allows multiple participants to train a model locally and share the model parameters. This approach reduces the risk of data leakage. To improve privacy protection, researchers have proposed a range of techniques that are designed to preserve privacy. These include differential privacy, homomorphic encryption, and secure multi-party computation, which enhance privacy protection by operating at different levels. Nevertheless, further research is required to achieve a balance between privacy and model performance in the context of FL. The present paper commences with an exposition of the notion of FL, clarifying its definition and rationale. Subsequently, a comprehensive review of extant architectures and classifications of FL is presented. The subsequent discussion focuses on the root causes of privacy threats in FL and analyses the risks that may be caused by data sharing and other links. On this basis, the advantages of privacy-preserving techniques such as differential privacy, homomorphic encryption, and secure multi-party computation in combination with FL are described in detail. These advantages include enhanced data security and privacy protection. The limitations of these techniques are also discussed. Finally, it comprehensively discusses the challenges of privacy protection in FL, such as the contradictory nature of ensuring both high model accuracy and efficient algorithms and the lack of unified quantitative standards.…
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.