Scholarlink

<p>The 6G networks have made the need to ensure privacy of the user data as well as high-performance communication urgent. The centralized machine learning models possess several issues such as high latency, privacy breach and non-scaled performance especially applied on dynamic distributed systems such as Human Activity Recognition (HAR). To address these shortcomings, the paper suggests a new framework would integrate Federated Learning (FL) with Edge Intelligence (EI) and Differential Privacy (DP). The proposed model allows decentralized activation of activity recognition models in edge devices themselves, thereby eliminating the need to transfer raw data to the central servers. Differential Privacy boasts an extra level of privacy since it adds a controlled noise to the gradient updates hence protection against data reconstruction attacks. The model was tested on the UCI-HAR data with 5-fold cross-validation and achieved impressive. It had an average accuracy of 99.25, an F1-score of 0.9871, and a Kappa score of 0.9909, which is a strong predictive performance and consistency. Also, the False Negative Rate (FNR) and False Positive Rate (FPR) decreased to 0.0027 and 0.0115, respectively, which indicates the accuracy of the model. Unlike other baseline models like SVM and CNN-BiLSTM, the given system significantly reduces the latency and communication overhead at a high accuracy and privacy guarantees. It can thus be used well in real-time smart cities, healthcare, and industrial IoT. In general, this research advances safe and effective edge-based AI schemes towards 6G systems. Federated Learning, Edge Intelligence, and Differentiated privacy combined to create secure, decentralized activity recognition is an innovation. It is a very accurate method that keeps the user privacy, reduces latency and lets the real-time processing operate without showing raw data beyond the 6G network.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Research on privacy protection and performance optimization of 6G communication network based on the fusion of federated learning and edge computing

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2025年12月13日

查看全文
  • Haifeng Zhao

摘要

The 6G networks have made the need to ensure privacy of the user data as well as high-performance communication urgent. The centralized machine learning models possess several issues such as high latency, privacy breach and non-scaled performance especially applied on dynamic distributed systems such as Human Activity Recognition (HAR). To address these shortcomings, the paper suggests a new framework would integrate Federated Learning (FL) with Edge Intelligence (EI) and Differential Privacy (DP). The proposed model allows decentralized activation of activity recognition models in edge devices themselves, thereby eliminating the need to transfer raw data to the central servers. Differential Privacy boasts an extra level of privacy since it adds a controlled noise to the gradient updates hence protection against data reconstruction attacks. The model was tested on the UCI-HAR data with 5-fold cross-validation and achieved impressive. It had an average accuracy of 99.25, an F1-score of 0.9871, and a Kappa score of 0.9909, which is a strong predictive performance and consistency. Also, the False Negative Rate (FNR) and False Positive Rate (FPR) decreased to 0.0027 and 0.0115, respectively, which indicates the accuracy of the model. Unlike other baseline models like SVM and CNN-BiLSTM, the given system significantly reduces the latency and communication overhead at a high accuracy and privacy guarantees. It can thus be used well in real-time smart cities, healthcare, and industrial IoT. In general, this research advances safe and effective edge-based AI schemes towards 6G systems. Federated Learning, Edge Intelligence, and Differentiated privacy combined to create secure, decentralized activity recognition is an innovation. It is a very accurate method that keeps the user privacy, reduces latency and lets the real-time processing operate without showing raw data beyond the 6G network.