Federated learning – a useful privacy-enhancing technique?
Federated learning is a machine learning and data science technique used to train algorithms without centralised data collection. Alessandra Calvi of Brussels Privacy Hub and Gianclaudio Malgieri (EDHEC/VUB) report.
Federated learning is a robust machine learning model which works without sharing data. The training data are kept on the device, making this technology more privacy friendly than centralised data collection.
It averts the privacy issues deriving from the transmission of personal data from a device to a centralised data collector, at the same time limiting the use of resources in terms of power and bandwidth. Practical applications range from improving software of virtual keyboards that perform predictive sentence completion, to providing aggregated information on the spread of diseases to health researchers.
International Report subscribers please login to access the full article.
If you wish to subscribe please see our subscription information.