The results of this component are intended to inform both technical users and possibly policy-focused audiences on the practical and efficient ways to implementing privacy-preserving machine learning using differential privacy.
The prototype and supporting materials will be shared through open-source repositories and technical documentation.
Addressing target audiences and expressing needs
- Use of research Infrastructure
Users who we can work with and provide them with the knowledge of how Differential Privacy can help with their data, and how it can help them share their data in a privacy preserving manner, helping them find utility to their data in a safe and GDPR compliant manner.
- Public or private funding institutions
- Research and Technology Organisations
- Academia/ Universities
R&D, Technology and Innovation aspects
The project is at a functional prototype stage, with core components tested on synthetic data. Software supports configurable model training and privacy parameter adjustment, providing a flexible foundation for privacy-aware machine learning workflows.
Next steps include:
- Validation across additional use cases
- Refine integration points for external systems
- Engage with partners for them to use software
The code can be made available on request, for people to run experiments using datasets that have been used during its development.
