PRIVACY-AWARE FEDERATED LEARNING ALGORITHM FOR HEALTHCARE IOT USING OPTIMIZED VGG16 WITH NOISE ADJUSTMENT

As healthcare technologies quickly improve, data protection and security have become very important issues. This is especially true as more and more Internet of Things (IoT) devices are added to healthcare systems. Federated learning (FL) seems like a good way to deal with these issues because it allows for decentralised model training while keeping private patient data kept locally. Existing shared learning algorithms, on the other hand, often have problems like models that don't work well together, data leaks, and threats from other computers. This article suggests a Privacy-Aware Federated Learning (PAFL) algorithm for Internet of Things (IoT) systems in healthcare. It uses an improved VGG16 deep learning design and a noise adjustment method. The suggested method improves the safety and performance of healthcare IoT apps by making sure that private data is never sent between devices. This way, models can still be trained on decentralised datasets successfully. Our method uses a noise adjustment system to hide gradients while the model is being updated. This lowers the risk of private information getting out and makes the model more resistant to possible hostile attacks. Because VGG16 has been optimised, model convergence is more accurate and efficient. This makes it a good choice for healthcare settings where making decisions quickly and in real time is important. The suggested Privacy-Aware Federated Learning algorithm does better than other federated learning methods in protecting privacy, making accurate models, and using computers more efficiently. This is especially true in healthcare IoT devices that don't have a lot of resources.