Secure Smart Health Care Data Transfer in Consumer Electronics Using Split Learning-Based Binary Sequential Split Vanilla Network

Split learning (SL) is a privacy-focused method for training deep learning models that prevents clients and servers from sharing raw data. There are a few downsides to supervised learning, despite its numerous advantages. The need for large client-side computing resources and the possibility of data privacy compromise are two major drawbacks. Binary Sequential Split Vanilla networks are suggested in this research as a way to safely and efficiently transfer data. In addition, the proposed fix may lessen the impact of accuracy while reducing privacy breaches caused by SL damaged data. Here, each client is tasked with training a subset of the model, called a "cut layer," that is smaller than the whole. The results of these cut layers are then sent to a central server, which continues the training process without having direct access to the initial client data. As an additional measure to supplement the present model and fortify privacy protection, it is also proposed including the bullet fish algorithm. This research tries to evaluate several benchmark models in relation to experimental findings by using the MUsculoskeletal RAdiographs (MURA) datasets. Lightweight applications in the IoT and Consumer Electronics sectors, particularly those involving mobile healthcare and other areas with high privacy protection concerns, may benefit from the suggested models, according to our research.