Faculty's privacy enhancing federated learning solution

点击翻译按钮获取中文摘要

Forecasting infectious disease to predict the behaviour of outbreaks, seasonal epidemics and global pandemics is rapidly becoming an essential part of a modern public health response. Modelling infection risk is highly complex and influenced by a range of epidemiological, socioeconomic and mobility factors and access to large scale public health data is essential to accurately model evolving disease dynamics. However, due to the entirely valid sensitivities regarding the privacy of such data and the resulting privacy regulation, accessing sufficient data for modelling can present a challenge. Faculty are developing a powerful and safe solution to address one of the most common challenges facing machine learning to enable privacy preserving federated learning. This solution will empower individual organisations, as small as individual hospitals or primary care practices, to benefit from a shared machine learning model that harnesses the benefits of large scale data from multiple organisations without the need for organisations to ever share their data. One of the main challenges to delivering high performance machine learning is the inability to facilitate model training on sufficiently large scale, real world, datasets as patients and organisations often have understandable privacy concerns. Faculty's federated learning solution will collaboratively train on privatised synthetic representations of individual organisations locally held data, thus avoiding GDPR or HIPAA regulatory hurdles, and securing sufficiently large scale data. This solution not only strengthens data privacy through privacy guarantee, but grants each contributing organisation the power to control the extent of their privacy vs performance pay offs and set the controls independently of both the central organisation and of each other contributing organisation without any raw or synthetic data transfer.