Deepfake detector
Definition
A machine-learning classifier trained to distinguish authentic recordings from AI-synthesised or face-swapped media. Outputs a probability score rather than a binary verdict. Published false-positive and false-negative rates are central to interpreting the score in a report.
Related terms
- C2PA (Coalition for Content Provenance and Authenticity)
- An open technical standard that embeds cryptographically signed provenance assertions into media files at the point of capture or editing. A C2PA...
- Error Level Analysis (ELA)
- A technique that re-compresses a JPEG at a controlled quality setting and maps the pixel-level difference between the re-compressed and original images....
- False-positive rate
- The proportion of authentic media files that a detection tool incorrectly classifies as manipulated or synthetic. Deepfake detectors and ELA tools both...
- Hash verification
- The process of computing a cryptographic hash (SHA-256 or equivalent) of the exhibit file and comparing it against a previously recorded value...
- Hedging language
- Qualified phrasing that accurately conveys the degree of certainty a method supports. Examples: 'consistent with', 'the findings are indicative of', 'cannot be...
Explained in
- Writing a Media Authenticity Examination ReportA machine-learning classifier trained to distinguish authentic recordings from AI-synthesised or face-swapped media. Outputs a probability score rather than a...