Reliable Federated Learning With GAN Model for Robust and Resilient Future Healthcare System

Federated Learning (FL) enabled the reliability and robustness of 5G communication networks for wireless edge computing to provide collaborative Deep Learning (DL) of complex models while protecting privacy for healthcare systems. Wireless end devices are more susceptible to corruption due to the vulnerability offered by open network settings, however, this creates security issues and lessens the effectiveness of DL-based security models for healthcare systems. Furthermore, disaster reliability in communication networks has garnered unprecedented attention from governments and companies, particularly during the current COVID-19 pandemic scenario. In this work, a novel reliable personalized Federated Learning-based Customized Inequality-Aware Federated Learning (CusIAFL) technique is proposed for securing color images while communicating with a wireless network. The proposed technique adjusts each data sampling to the local target during optimization using knowledge of client-label availability. The work that is being presented uses a hybrid technique to maintain consistency in the time-series data. and a novel Pix2Pix Generative Adversarial Network (GAN) technique is used to generate realistic images. This novel work is tested on different non-medical and medical images. The experimental results have been evaluated using performance metrics, namely accuracy, entropy, PSNR, HD95, SSIM, and MSE. Furthermore, the accuracy varies from 89 to 93 percent with different datasets outperforming well with existing SOTA techniques. The outcomes demonstrate that the proposed CusIAFL-based scheme is more effective than the State-Of-The-Art (SOTA) models.