2
\$\begingroup\$

I'm currently working on some software to help identify fraudulently attributed images or ones that have been sent in with altered attribution.

As such, I'm interested in working out which EXIF fields (or potentially, any other data points) could be used to identify the creator or the device which took the picture, especially compared to others in a set. Often times, a batch of images come in with a mixture of metadata (but look as though they may be from a similar artist/user).

There are obvious ones like Author and Make/Model, but I wondered which other fields contain unique information - or at least ones that could potentially identify picture A as not being taken by the same device as picture B was taken with.

\$\endgroup\$
2
  • \$\begingroup\$ What makes you so certain a photographer only uses one device? Most of us shoot with a phone and a camera, if not multiple cameras. \$\endgroup\$
    – inkista
    May 16, 2017 at 18:02
  • \$\begingroup\$ @inkista For the project I'm doing, I'm mostly interested in assessing if the device is the same, not the photographer. For instance; user states picture A and B were taken with Camera X, if B doesn't look to be taken with the same device, I can always ask for another image to be taken with Camera X to confirm Camera X is in their possession. \$\endgroup\$
    – Retro
    May 17, 2017 at 23:03

1 Answer 1

2
\$\begingroup\$

A few things to think about.

  • If you aren't limited to Exif, don't be. IPTC/XMP may contain additional fields.
  • You may be able to use information that identifies the camera (eg. serial number) to correlate images from the same device, as that strongly suggests the same author.
  • Some cameras store fairly detailed information about the lens too. Two images with matching lens data are more likely to come from the same author.

Much of this extra stuff is in the makernote, so you'll need a tool that can get that out for you too. What software are you using? If you're coding in C# or Java (or other CLR/JVM languages), check out my metadata-extractor project.

For some real world examples, check the metadata subfolders of specific file formats in this test data collection. It may prompt some further ideas.

\$\endgroup\$
4
  • \$\begingroup\$ Thanks for the response, it's all really helpful! I'm currently coding in Python and I'm also bringing in IPTC data - XMP data is more difficult to get hold of without bringing in a lot of dependencies. I guess my follow on question is what identifying fields (other than serial number) are worth looking at? \$\endgroup\$
    – Retro
    May 10, 2017 at 16:28
  • \$\begingroup\$ You may be able to call out of process to exiftool. \$\endgroup\$ May 10, 2017 at 16:47
  • \$\begingroup\$ I'm currently grabbing the exif data with the 'exifread' module which works quite well, my issue is which fields are useful in identifying the same device. \$\endgroup\$
    – Retro
    May 10, 2017 at 18:09
  • 1
    \$\begingroup\$ I don't have anything more to add regarding fields. However note that exiftool supports all the metadata under the sun, not just Exif. \$\endgroup\$ May 10, 2017 at 23:00

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.