Scholarlink

A challenge of primary concern in healthcare today is that hospitals have to secure the extremely critical information of their patients, which arises through the Internet of Medical Things (IoMT). A combined federated learning (FL) framework and Differentially Private Generative Adversary Networks (DP-GANs) are proposed in this study to improve privacy, scalability, and performance in IoMT environments. The architecture consists of three layers: data preprocessing for the device level, edge-level model training with secure gradient aggregation, and cloud level optimisation. DP-GANs provide data synthetization that preserves statistical utility over protecting the privacy of persons in the sense of exposure to adversarial risks and threats in re-identification. It has been found to generate data that preserves statistical utility synthetically. The results show that the proposed framework provides nearly 0.95 accuracy and 0.93 precision with total privacy, thereby vigorously protecting the confidentiality of information when using MIMIC-III Healthcare Data. This research highlights the possibility of introducing state-ofthe-art methodologies that preserve privacy within FL to protect highly sensitive healthcare data while retaining the utility of models applicable in building robust, scalable, and secure IoMT applications.

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

Safeguarding Patient Privacy: Leveraging Federated Learning for Enhanced Anonymity and Identity Protection in Healthcare

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Tanushka Guleria,
  • Naveen Chauhan

摘要

A challenge of primary concern in healthcare today is that hospitals have to secure the extremely critical information of their patients, which arises through the Internet of Medical Things (IoMT). A combined federated learning (FL) framework and Differentially Private Generative Adversary Networks (DP-GANs) are proposed in this study to improve privacy, scalability, and performance in IoMT environments. The architecture consists of three layers: data preprocessing for the device level, edge-level model training with secure gradient aggregation, and cloud level optimisation. DP-GANs provide data synthetization that preserves statistical utility over protecting the privacy of persons in the sense of exposure to adversarial risks and threats in re-identification. It has been found to generate data that preserves statistical utility synthetically. The results show that the proposed framework provides nearly 0.95 accuracy and 0.93 precision with total privacy, thereby vigorously protecting the confidentiality of information when using MIMIC-III Healthcare Data. This research highlights the possibility of introducing state-ofthe-art methodologies that preserve privacy within FL to protect highly sensitive healthcare data while retaining the utility of models applicable in building robust, scalable, and secure IoMT applications.