10 November 2020
Press coverage on Machine Learning Privacy Meter
A team from N-CRiPT, led by NUS Presidential Young Professor Reza Shokri, has developed a tool that measures the privacy risks of machine learning models. The tool, named Machine Learning (ML) Privacy Meter, enables data analysts and researchers to assess the extent to which their models leak information about their training data by running inference attacks against the models.
The tool was released publicly on GitHub and has been incorporated into Google’s developer toolkit to test the robustness of its AI algorithms’ privacy preserving features.
Moving forward, Professor Shokri is leading efforts to partner industry in exploring the integration of the ML Privacy Meter into AI services. His team is also building algorithms for privacy-preserving AI model training.