Secure Federated Learning Approaches to Diagnosing COVID-19

Our paper proposes building a model through federated learning to automatically diagnose COVID from patient chest X-Rays The recent pandemic has delivered a desire for understanding efficiently and accurately. This model would aid as a useful supplement COVID-19 diagnoses among patients in hospitals. Currently, hospitals to doctors who are trying to quickly determine whether have difficulty diagnosing this novel respiratory illness from a patient needs to continue with the RT-PCR test from their chest a patient's chest X-Ray to another, simply because it is difficult X-Ray results. Initially, one might question why the model needs to compare COVID chest X-Ray between patients due to HIPAA to be trained in a federated setting - wouldn't it be enough to train compliance. In this paper, we aim to build a model to assist in the separate models per hospital? However, our concern is that the data diagnosis of COVID-19 while being HIPAA compliant through federated at a single hospital may be severely biased and would not be an learning, a distributed machine learning technique used to accurate representation of all patient chest X-Rays across hospitals train an algorithm across multiple decentralized devices with local [33]. Thus, we utilize federated learning techniques to ensure that data samples without sharing them [8]. Our model extends on existing the model will be able to generalize well to samples from the other work in the chest X-Ray diagnostic model space; we analyze hospitals. Additionally, since it is inherently possible for an attacker the best performing models in the CheXpert [15] competition and to learn personally identifiable information from a model, as can be build our own models that work effectively for our hospital data.