Federated Learning is a new Google initiative to improve its machine learning model while offering greater protection of users’ privacy.

Machine learning sans snooping

News stories about data security, integrity and intelligence have motivated many service providers to beef up perceptions about how they handle privacy.

Google has long used search data, user behaviour and device usage stats as a basis for service improvements and refinements. Now, with Federated Learning, they’re on to a way to achieve the same ends with less identifiable data needed.

How it works

As Google reports, each device will download the current code from a central server. Then, it will formulate suggested improvements using data on the device.

Each device summarises its suggested changes to the code as a ‘small focused update’. That update is encrypted and sent to the cloud.

Branding by privacy perception

Increasingly, for online service providers, protecting user privacy is a matter of positive branding.

Projects like Federated Learning are a sign of good faith from an all-but ubiquitous tech giant. This model shows they can pursue their crowd-sourced-data-driven process of improvement while remaining committed the concerns of their market.

Will these changes affect your work? Let us know in the Comments.

Copy Transmission is a Melbourne-based agency :: Better Brands. Loud & Clear.