Building a User Data Protection Model by Combining Distributed Storage and Machine Learning

Abstract

With the acceleration of the digitalization process, the risk of user data leakage has become increasingly prominent. Especially in fields such as healthcare and finance, the leakage of users' sensitive information may lead to serious consequences. This study aims to combine distributed storage with machine learning to build a new user data protection model. Therefore, this research will adopt federated learning as the framework, integrate lightweight deep learning models with differential privacy technology, to realize decentralized data storage and privacy-enhanced model training. This approach is intended to address the issues of traditional machine learning's reliance on centralized data and the hidden risks of privacy leakage. In addition, this study proposes a "edge-region-central" hierarchical architecture, a three-level privacy protection system, and a storage-computation collaboration mechanism. Finally, the advantages of the model in terms of privacy protection effect, accuracy, and efficiency will be verified through experiments, providing technical support for user data security.