Multimodal AI security and provenance
Trace trust across every input.
How can AI assess security and evidence of origin when information moves across text, image, audio and video?
Multimodal systems combine inputs whose origins and trust levels may differ. We explore how that combination affects model decisions, how transformations change the evidence available and what provenance can establish about content without overstating its authenticity.
Research areas
Study adversarial inputs and conflicting signals across text, images, audio and video.
Examine trust boundaries when a system converts or combines different media.
Investigate content provenance, attribution and the limits of authenticity signals.
Evaluate security and provenance methods under ordinary edits and unfamiliar generators.
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.