Camera vendors are perfectly capable of storing a lot of metadata in the jpeg encoded image itself, in an steganographic way. If you want to really be sure, crop your image to a position that is not multiple of 8 on any direction, then scale it by a factor very close to 1, add some noise, and re-compress it again. Some steganography is yet robust to that, but only the really fancy stuff.

Anyhow, if your image shows recognizable landmarks with shadows, then it will be feasible to recover the exact point of view and the time of acquisition.

Use image pool to train a GAN, publish the GAN images instead of the real ones.

> Anyhow, if your image shows recognizable landmarks with shadows, then it will be feasible to recover the exact point of view and the time of acquisition.

With what level of accuracy? Are you claiming you can confidently assert 2020-12-21 15:12 over, say, 2020-12-22 15:15 via shadows

> confidently assert 2020-12-21 15:12 over, say, 2020-12-22 15:15 via shadows

Not at all. By visual inspection you can see whether it's the morning or the afternoon (if you know the orientation of the landmarks). I guess by image processing you could get a precision of +- 1h maybe?

How precise is a sundial? How high is the resolution of the camera? I'd imagine that on a bright sunny day with a high resolution camera showing clean sharp shadows, it'd be possible to get a precision much better than +/-1h.

Sure, I was thinking about a photo of human subjects in the foreground, with some distant buildings in the background. But if you photograph the shadow of the Eiffel tower on a clear sunny day you may be down to the precision of a couple of minutes.

But that's only if you can resolve the "which day" issue?

Oh, yes, that seems a hard problem. But with sufficient effort (cloud patterns, parked cars, other people in the image) there is still a lot of information to extract. Especially if other people were taking exim-tagged photos in the same time and place and putting them on instagram.

The arc of the sun through the sky, along with observable weather and variability in air quality could let you pin the date down with surprising accuracy. Two neighboring dates with similar weather would obviously be very hard, but even variation in cloud patterns could be enough to make an id.

The problem with that idea is that there would have to be a stenography standard and then it becomes easy to defeat.

Of course, the idea is to keep it secret; a "stenographic standard" seems absurd. If they are embedding this information, it will not be readily detectable by most people.

iOS now have an option to do this natively. But I find it bit convoluted. My go to app for stripping out metadata from photos is Exif Viewer.

My camera apps never include GPS coordinates to begin with (which is enough stripping for me, but then again I'm not part of Qanon and all that).