Is privacy in the age of AI (artificial intelligence) a lost cause? One new research program aims to create greater privacy, aka privacy by design, where the AI platform never sees a face on an image, but the final edited image still looks natural.
The research is being done at Purdue University, where a new patent-pending system leverages before and after photos uploaded to an AI editing platform. It allows users to mask sensitive regions of a photo, such as the face. The technology can then reintegrate the sensitive region back into the image.
Here is how this can help:
- Full privacy, as sensitive data never leaves the user’s device.
- High-quality edits and photorealism.
- Compatibility with commercial generative AI models.
Research shows the system reduced the ability of AI models to detect attributes such as eye color, facial hair, and age group. In some cases, attribute-classification accuracy dropped by more than 80%, demonstrating strong protection against identity leakage.
Looking to the future, the team is taking steps to bring the patent-pending technology to the real world. Researchers hope to expand the system to protect additional sensitive features such as medical details, ID documents, and other privacy-critical content.