FAQ 21 - Proof

  • 2101. What ways do TTG photographers have of convincing viewers that a photo qualifies for the TTG label?

    That is covered in the the background brief on “convincing the viewer”. (Summary: clearly distinguish TTG images from non-TTG images, save originals, save neighboring frames, save supporting images, and save exposure data.)

    If Verifiably Unaltered Originals (discussed below in #2103 and following) become viable, they could help photographers prove when a photo is undoctored.

    But “being undoctored” is only one component in the package of things that determine a photograph's trustworthiness (see also {x17} below).

  • 2102. Will there ever be a way for the typical photographer to prove beyond all doubt to everyone that a photograph is qualified for the TTG label?

    No, there never will be a way to prove that — not “beyond all doubt,” not to “everyone.”

    There are endless variables involved, with respect to both the seen aspects of photographs (the “appearance,” which P7 addresses) and with respect to unseen aspects of photographs (the “circumstances,” which P8 addresses).

  • 2103. Could “content authentication” technology be used to prove that a photo is undoctored?

    Yes. Comparing a TTG-labeled photo to the unaltered original version of that same photo would reveal whether the photograph meets TTG’s definition of “undoctored.”

  • 2104. And the “before” version of those “before vs. after” comparisons is what TTG calls a VUO?

    Yes, TTG’s term for an encrypted unaltered original is a “VUO,” or a verifiably unaltered original.

    TTG’s term for the VUO that has been verified as being unaltered is a “CUO,” or a confirmed unaltered original.

    Both of those terms were created by TTG in 2021 in anticipation of future technologies. If a different term comes into popular use, these references will be changed accordingly. (See also the note at the bottom of this page.)

  • 2105. How would the comparison be made between the Confirmed Unaltered Original and the final, TTG-labeled photo?

    It would be a visual comparison, using TTG’s Allowable Changes list coupled with the Trust Test, to confirm that no disqualifying changes were made to the photo.

    It would be easy in any photo-editing program to make a layer of the CUO “before” picture and a layer of the TTG-labeled “after” picture and then click either layer on or off to see what was changed.

  • 2106. What's to keep a photographer from submitting a Verifiably Unaltered Original that is actually a rephotograph of a doctored photo?

    The VUO would have embedded in it the focus data, location data, the time data, and the device data.

    The focus data in most cases would reveal whether it was a photograph of another photograph.

  • 2107. What will it take for the VUO process to be widely accepted?

    Once “trust” considerations are definitively resolved (“Is the encryption process truly unhackable?”), to gain widespread acceptance the procedures for both encrypting and for verifying will have to be both user-friendly and inexpensive (or free).

  • 2108. What about the technologies that reveal exactly what changes were made to a photograph after it was encrypted at the time of recording?

    They can be helpful for some image providers, but that information isn’t necessary for TTG purposes.

    All that’s needed for TTG purposes is a “before” photo that can be verified as being unaltered and contains the exposure data noted in #2106 above.

    A simple visual comparison of that unaltered original with the TTG-labeled photo would reveal whether the photo meets TTG’s definition of “undoctored.”

  • 2109. And neither the “Verifiably Unaltered Original” nor the “Confirmed Unaltered Original” would be created to be put before viewers?

    That is correct.

    Since even minimal cropping or downsizing – both of which are routine even in news photography – could keep a VUO from being confirmable as a CUO, it would be unrealistic in most contexts to expect the VUO or CUO version of a photograph to be put before viewers.

    (The CUO is the same image as the VUO and thus the two look identical. The only difference is that the CUO is a confirmation that the VUO’s encryption is intact.)

  • 2110. But a version of the VUO/CUO (“before”) photo could be published alongside the TTG-labeled “after” version to help resolve a controversy?

    Yes, after taking precautions against image theft, a version of the VUO/CUO could be put before viewers if needed to resolve a controversy...

    . . . the way photo-contest sponsors sometimes publish a copy of the unaltered original file to explain why they disqualified as “doctored” a particular photo submitted to the contest.

  • 2111. How is “unaltered original” defined?

    For standalone cameras, what constitutes the “unaltered original” of a single-exposure photograph would be akin to a typical RAW image.

    But it may be harder to reach an industry consensus on what constitutes an “unaltered original” for smartphones. Unaltered originals of individual exposures on smartphones can be fairly unimpressive before they are combined with other exposures and extensively massaged—

    —that is, they would be unflattering to the smartphone maker.

  • 2112. Photojournalists can already buy cameras with authentication technology. Isn't that enough?

    The issue is that more and more spot-news photographs are being supplied to news orgs by ordinary citizens with a smartphone — not by photojournalists using a high-end standalone camera.

    AIFI technology is making it ever easier to doctor images instantly on smartphones. Thus it could be argued that to a news provider, it would be more useful to be able to verify the undoctored status of “smartphone photos that were submitted from strangers” than to verify “standalone-camera photos taken by established photojournalists that are known and trusted.”

  • 2113. Could VUO/CUO technology be built into smartphones?

    Yes. Smartphones may be even better suited for implementation of VUO/CUO technology than interchangeable lens (standalone) cameras are, for two reasons:

    1. Smartphones are more likely than standalone cameras to be connected to the web/cloud, which could help with quickly uploading encrypted VUOs.

    2. Smartphones have “sealed” lenses that cannot be interchanged. Lens/focus data is critical in image verification — especially to prevent rephotographing doctored or aigmented images — but many standalone camera photographers use interchangeable third-party lenses that may not provide reliable focus data.

    (Accessory lenses can be attached to smartphones, but it would be easy for manufacturers to position next to phone lenses a sensor to detect anything within a couple of inches of the lens. Focus-data discrepancies would often also reveal the use of accessory lenses.)

  • 2114. But if unaltered originals can make smartphones look bad (as per #2111 above), why would smartphone manufacturers ever integrate VUO/CUO technology into their devices?

    Because of demand.

    As we speed into the age of AI, it may not be long before in many situations the “trustworthiness” of images is as important as the “appearance” of images.

    (Up to now, smartphone manufacturers have been spending billions of dollars a year on “appearance” and virtually nothing on “trustworthiness.”)

    No major smartphone manufacturer is likely to want to be known as the maker of “the phone that takes photos nobody trusts” (a danger with smartphones marketed for their AIFI-manipulation capabilities).

  • 2115. Will it eventually be possible for anyone to do the verification process and not just newsphoto editors?

    For example, would a photographer who puts a TTG-labeled image on social media be able to say, “Check the CUO and confirm for yourself that this image is undoctored”?

    Perhaps, but there will always be the issue of preventing image theft if the CUO is made available to the wider public.

    There are workarounds to image theft, such as prominent watermarking and/or posting a low-res version of the CUO.

    But those workarounds would obviously compromise the verifying persons’ ability to see a non-watermarked, high-resolution CUO — which is the best comparison option when assessing a high-resolution TTG-labeled photograph.

  • 2116. What are the details of how VUOs and CUOs would actually work if implemented on millions of smartphones?

    VUO/CUO definition

    Would the encryption be a standard feature on new smartphones? • Would it be backwards compatible with existing smartphones? • Would photographers have to install new software to use it? • Who will control the software standard? • Would the encryption/verification process be universal or would it be specific to each brand of smartphone? • Would it cost extra to encrypt, upload, and store VUOs compared to normal photo uploads to storage services? • How much does the encryption add to the file size? • What’s involved in confirming that a VUO’s encryption is intact? • Since verification involves access to a full-sized original photo, who would be trusted to confirm VUOs as being unaltered? • What would be the protection against false positives or false negatives? • Who would be allowed to examine the VUO/CUO?

    Those issues will have to be figured out by companies that bring the VUO/CUO concept to market.

  • 2117. Will VUO/CUO technology make TTG obsolete?

    No, there’s no chance of that.

    A VUO/CUO can only prove whether a photo is undoctored, but as the Trust Test demonstrates, being “undoctored” is only one of many trust-related considerations for photographs.

    More

    • A note for this FAQ only

  • Because this FAQ includes speculation about a technology that is not yet widely available, unlike with the other FAQs on this website the content and order of multiple questions and answers above will likely change in the future.

    Thus users can make a link to any specific question in the usual way, but they should be aware that (unlike with the other 19 FAQs) the questions and answers in this FAQ (above) may change.