PAT-AI: Privacy-Aware Transparency for AI in Singapore

Project Description
Recent years have seen an increasing push to make black-box algorithmic systems more transparent to various stakeholders. However, making systems more transparent runs the risk of making them more exposed to information leaks, e.g. private user information can be inferred via transparency reports. In this project we study how one might mitigate the effects of such leaks, and design novel transparency measures that are resistant to adversarial inference attacks.
Research Technical Area
- Game theory and economic paradigms
- Machine learning
- Privacy & transparency
Benefits to the society
Project’s Publications
–