FL-DRNA: Federated Learning with Decentralized Record Network based Auditing

The proliferated utilization of Artificial Intelligence (AI) has been emerging high in healthcare sector which has introduced potential benefits in various form while it has induced critical security challenges at same time. Specifically, the problem of data privacy is on risk for multi-chains hospital or any healthcare units whose data is maintained in distributed system. Therefore, this paper presents an innovative framework known as Federated Learning with Decentralized Record Network based Auditing (FL-DRNA) which is meant for addressing this problem of data privacy without sharing any form of sensitive information of patient. A collaborative model training is facilitated by proposed model connecting multiple healthcare units while Advanced Encryption Standard (AES) is used for preserving the privacy during communication as well as during validating the transactions. The model unifies decentralized auditing with adaptive federated aggregation for improving communication integrity, trust validation, and transparency. Adopting standard benchmarked MIMIC-III dataset, the outcome of the FL-DRNA is noted to be significantly better in contrast to existing standard approaches. FL-DRNA reports of contributing to 97.42% of accuracy while it reports 96.83% of trust validation score, which is notably higher in contrast to baselines.