Detect
Identify AI-generated images and video, deepfakes, and content manipulations from editing software so that you know what’s real.
Media authentication grounded in forensic science beyond binary decisions
Unsurface detects manipulated and AI-generated images and video, identifies altered regions, and explains the forensic evidence behind its conclusions. Our technology builds on more than a decade of peer-reviewed media-forensics research, including foundational work from DARPA’s MediFor & SemaFor programs. We serve organizations making high-stakes decisions where the authenticity of visual media matters.
Forensic analysis 2 regions
Verdict
Falsified image — splicing
This is a falsified image where two women standing on the right has been spliced into a photograph of two men at an outdoor venue.
Our analysis shows that the two men on the left were captured by a Sony ILCE-6000, whereas the inserted subjects on the right originate from a Nikon D3200, indicating two different cameras were used to create this composite. Furthermore, the spliced group was downscaled to approximately 77% of its original size and compressed at JPEG quality ~91, while the host scene remains unresampled with a lower compression factor of ~85.
Localize
Pinpoint the image region or video frames containing manipulated or AI-generated content so that you know where media is falsified.
Attribute
Determine the likely camera, AI-generator, or editing tools used to falsify media so that you know how it was faked or modified.
Explain
Combine forensic findings into explainable evidence that analysts can inspect, interpret, and use to support high-stakes decisions.
How it works
Unsurface’s media authentication technologies work by exploiting forensic traces left in an image or video, similar to how a criminal leaves behind fingerprints at a crime scene. These traces are minute statistical relationships between pixels that are left in an image or video by the camera that captures it, the AI generator that creates it, or the editing software that modifies it.
Unsurface’s technical approach has been validated by dozens of peer-reviewed publications in top-tier academic venues and has its roots in DARPA research under the MediFor and SemaFor programs. Early versions of our core technology have already been deployed in government systems and used to authenticate images and videos for several news agencies in high-profile stories.
Request a demo
We work with a limited number of founding partners to shape Unsurface for high-stakes verification workflows. If your team makes decisions that depend on the authenticity of visual media, we'd like to hear about your use case.