Federated learning as an analytical framework for personal data management – a proposition paper

Data minimisation and storage limitation are two principles incorporated in the GDPR aimed to increase personal data subjects' control over their own data and put restrictions on the amount of information that may be extracted from them in the data mining process. Implementation of those two principles has always been a challenging task, as their interpretation is discretional and current legislative measures may not necessarily protect data subjects adequately. In this paper, we introduce the concept of distributed learning as a viable tool for implementing data minimisation and storage limitation principles and argue that perhaps it could be appropriate to consider a branch of distributed learning, namely the concept of federated learning, as an analytical measure for guaranteeing data limitation and minimisation. To further support this thesis, we discuss how Federated Learning may be used in geospatial data analysis while the final outcomes of the experiments are yet to be published.