An Efficient Privacy-Preserving Federated Learning Scheme for Collusion Resistance and Dropout Prevention
Abstract
In the field of distributed machine learning, federated learning (FL) is used as a data privacy-preserving technique that allows multiple clients to collaborate in training a shared model without disclosing local data. However, FL faces the challenges of inefficient communication and high computational cost in practical applications. In this paper, we find that EPFL schemes have certain security flaws, which will result in the leakage of user privacy once the server colludes with the clients. To address the shortcomings of the EPFL scheme which is not anti-collusion and anti-dropout, this paper proposes an improved federated learning algorithm, which adopts the Flamingo protocol, and has the properties of anti-collusion and anti-dropout under the premise of ensuring data privacy, and has a low communication volume. A lightweight exit resilience protocol ensures that the server can obtain meaningful results even if the client exits in the middle of the process. The analysis results show that compared to the original scheme, this scheme has anti-collusion and anti-dropout features while maintaining the model's accuracy.
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.