Provenance Detection System for Deep Learning Content in Healthcare

Abstract

In this article we provide a general framework using Ethereum smart contracts to track back the provenance and evolution of deep learning content (DLC) to its original source even if the DLC was edited (e.g. DL models were retrained or/and datasets were updated) by anonymous authors. The main principle behind the solution is that if the DLC can be credibly traced to a trusted or reputable source, the DLC can then be real and authentic. The solution is proposed in the healthcare context and for medical DLC, especially for federated machine learning, but it can be applied to any other form of DLC.