Uoften

Uoften is a fictional open-source platform designed for experimenting with privacy-preserving personalisation in content delivery. The project provides a modular software stack intended for researchers and developers to prototype recommender systems that respect user privacy while allowing real-time and batch data processing.

Conceived in academic settings in the early 2010s, Uoften emerged as a collaborative effort among universities and independent researchers. The first public code release was in 2012 under a permissive license, with subsequent major revisions in 2016 and 2019 expanding the architecture to support federated learning and differential privacy adapters.

Its core architecture consists of a core engine, data adapters, model templates, and optional privacy modules. The core handles scheduling and orchestration; data adapters connect to data sources; model templates provide reusable recommender strategies; privacy modules enable differential privacy, secure aggregation, and on-device learning. The platform supports both real-time streaming and offline processing.

Uoften has been used by academic groups, non-profit research labs, and some startups to explore privacy-centric content delivery. Adoption metrics are limited due to its experimental nature, but users cite flexibility and interoperability as strengths, along with challenges in setup complexity and documentation.

Maintained by a distributed community, Uoften is released under a permissive open-source license. Governance is handled through community contributions and a lightweight project steering group, with decisions made by consensus and code of conduct guidelines.

See also: privacy-preserving machine learning, recommender systems, federated learning, differential privacy, open-source software.