Is there a way to verify that photos taken with a digital camera doesn't have any identifying invisible watermark?

Question

Digital cameras these days are capable of recording millions of colors, out of this many colors, I doubt humans can distinguish between hundred colors. So is it not possible for digital camera manufacturers to create a hardware or software which might add invisible watermark on the photos taken with a camera to identify that it was taken with a particular camera.

Is there a way to identify or remove such watermarks?

Not that I know of, but why would they? EXIF data exists for that same purpose; to identify manufacturer, model, firmware version as well as image details. — spikey_richie, Dec 24 '21 at 15:00
@spikey_richie But there are tools to erase or spoof the EXIF data but if it existed in the pixels, it would be much more difficult if not impossible to remove the information. A whistle blower might get caught if such information existed. — bypass restrictions, Dec 24 '21 at 16:47
@spikey_richie For the same reason that a lot of color printers include yellow tracking dots — Peter M, Dec 24 '21 at 16:47
I’m voting to close this question because it's really, really hard to prove a negative. No matter how much analysis you do, you still might have missed something. This probably belongs on Information Security, not here. It has nothing to do with photography, it's an element of steganography. — Tetsujin, Dec 24 '21 at 18:40
Somewhat related questions: Can digital cameras sign images to prove authenticity?, How can I tell if two pictures were taken with the same camera without metadata?, Can information like a signature, secret key or password be encoded in a photo?, and How are copyright infringements verified when a photo is initially made in JPEG, not raw? — scottbb, Dec 25 '21 at 00:53
The point of the last link in my previous comment is that one of the answers pointed out that cameras have "fingerprint"-like noise patterns that can be used to likely identify/verify if a particular photo came from a certain camera, much like bullet profiling to identify which gun fired the round. — scottbb, Dec 25 '21 at 00:55
This is actually closely related to the field of steganography. What you said about the millions of color is true; in fact, it's easily possible to hide entire images within another photo, far more information than a small watermark. However, I wouldn't be too worried about camera manufacturers doing this, and any sort of lossy compression, resizing, or editing would be likely to destroy the information anyway (and since those are so commonly done, it would not be worth it for a manufacturer to attempt that). — Radvylf Programs, Jan 24 '22 at 05:11

score 2 · Answer 1 · answered Dec 25 '21 at 00:09

Every sensor is unique. That's why they are screened for hot/dead/stuck pixels at the factory and pixel mapping is done to each one before they're sent to the marketplace to be sold. With the millions of photosites on modern sensors, the probability that two sensors, even from the same batch, have the exact same mapping is pretty much statistically impossible. It's every bit as unique as a fingerprint.

If someone has the specific camera in question in their possession, they can fairly easily verify if a photo taken with that specific model camera came from that specific copy or not. No "watermark" needed. It's already there in the raw data off the sensor. That's assuming the image in question is reasonably exposed so that it's not completely blown out with every single pixel fully saturated, in which case the raw data contains no real usable information of any kind.

It's also possible to verify or eliminate if two disparate images came from the same camera or not. This is a bit more dependent upon the scene contents and exposure levels of each image, but it's still possible to compare two images and determine whether they were taken by the same specific camera or not with a fairly high degree of certainty.

score 1 · Answer 2 · answered Dec 24 '21 at 16:06

1

Humans can see millions of colors (~11M), but most cameras can record/reproduce more (16M+)... so your premise isn't w/o validity.

In fact, there are programs/services for adding invisible watermarks to photos, and also for detecting those watermarks... e.g iMATAG.

But no camera manufacturer is automatically placing invisible watermarks on images; and I doubt they ever will; there is already plenty of normally invisible data in a digital image for those purposes. And one can easily create a watermark that would be excessively hard for another individual to see w/o knowing how to reverse the edit.

Besides, a watermark/EXIF/IPTC is not required to identify the original source/authenticy anyway...

answered Dec 24 '21 at 16:06

Steven Kersting

17,087
1
11
33

What are plenty of normally invisible data that can be used to identify? – bypass restrictions Dec 24 '21 at 16:46
@bypassrestrictions; given your user name, I don't feel inclined to say... I happen to participate in this group, and know these things, because I also happen to be a photographer; and I value my IP rights. – Steven Kersting Dec 24 '21 at 17:05
2

The assertion that no camera manufacturer is placing watermarks warrants a citation or restatement as conjecture based on your experience. – Bob Macaroni McStevens Dec 24 '21 at 18:04
1

Given that one of the major camera manufacturers was one of the inventors of the scheme used to put invisible watermarks on printer output, it is certainly a legitimate question as to whether they did the same to their camera output. – Philip Kendall Jan 24 '22 at 10:47
1

A manufacturer doing that kind of thing in a professional camera might lose a lot of goodwill from the photojournalist community if caught, or worse if someone gets hurt... wonder if they'd want to risk that! – rackandboneman Jan 24 '22 at 12:28

Is there a way to verify that photos taken with a digital camera doesn't have any identifying invisible watermark?

2 Answers2