It is concentrated at the top: the five leading assignees together hold 75.5% of all 106 records in scope, with the single leader accounting for 30 records on its own. Past the top five, counts drop sharply — fifth place holds 6 records and tenth place only 2. This means a new entrant is competing against a small group with substantial filing depth, but the steep drop-off after the top ranks also means there is room for narrowly-targeted filings that the leaders have not covered.

Tamper detection answers whether an image or video has been altered at all; tamper localization goes further and identifies which region of the file was manipulated. The patents in this dataset overwhelmingly target the localization problem, using signals like sensor pattern noise, compression history and resampling traces to pinpoint edited regions rather than just flag a file as suspect. That distinction matters for freedom-to-operate review, since a claim to a specific localization method does not necessarily block a simpler binary-detection approach.

G06F (electric digital data processing) covers 50.0% of the 106 records because many forensic detection methods are claimed at the level of general-purpose data processing — the algorithm, the model architecture, the data pipeline — rather than as image-specific hardware or optics. G06K, G06T and G06V, which are more imaging-specific, each cover smaller though still substantial shares. A record can and often does carry more than one of these classes, which is why the shares add up to well over 100% of the record total.