Deepfakes & authenticity
Assessing trust in synthetic media.
How can AI defenders assess authenticity when AI attackers create convincing synthetic media or impersonations?
This exploratory area examines how AI defenders assess audio, images and video used for impersonation or deception. Research questions connect media analysis with source history, provenance and the context in which a claim is made. Evaluation should test unfamiliar generators and realistic transformations, while making uncertainty clear to the person deciding whether to trust or verify the material.
Research areas
Combine signals within and across audio, images and video with contextual evidence about an impersonation or claim.
Evaluate whether a detector generalises to unfamiliar generators, compression, editing and changes in recording conditions.
Explore how provenance and source history complement content-based authenticity assessment.
Study how confidence, conflicting signals and verification options can support a human decision.
Publications & related resources
Projects, publications and research materials will be added here.
Contribute to this research.
Explore student research, internships and collaboration with the lab.