Technology

59414 readers

3769 users here now

This is a most excellent place for technology news and articles.

Our Rules

Follow the lemmy.world rules.
Only tech related content.
Be excellent to each another!
Mod approved content bots can post up to 10 articles per day.
Threads asking for personal tech support may be deleted.
Politics threads may be removed.
No memes allowed as posts, OK to post as comments.
Only approved bots from the list below, to ask if your bot can be added please contact us.
Check for duplicates before posting, duplicates may be removed

Approved Bots

founded 1 year ago

MODERATORS

288

‘There is no such thing as a real picture’: Samsung defends AI photo editing on Galaxy S24 (www.techradar.com)

submitted 9 months ago by leo@lemmy.linuxuserspace.show to c/technology@lemmy.world

109 comments fedilink hide all child comments

you are viewing a single comment's thread
view the rest of the comments

[–] NotMyOldRedditName@lemmy.world 1 points 9 months ago* (last edited 9 months ago) (2 children)

Are the raw photos manipulated or are they just the original 1s and 0s unedited?

Not sure if there's any preprocessing before the processing.

Edit: I'm imagining a digital camera that cryptographicly signs each raw frame before any processing with a timestamp and GPS location. Would be the best you could probably do. Could upload it's hash to a block chain for proof of existence as well.

Edit: I guess the GPS system would need some sort of ceyptographic handshake with the camera to prove the location was legitimately provided by the satellite as well.

[–] Kazumara@feddit.de 4 points 9 months ago

the original 1s and 0s

I think your issue starts there, you already have to decide how to build your sensor:

If it's a CMOS sensor how strong do the MOSFETs amplify? That should affect brightness and probably noise.
How quickly do you vertically shift the data rows? The slower the stronger the rolling shutter effect will be.
What are the thresholds in your ADC? Affects the brightness curve.
How do you layout the color filter grid? Will you put in twice as many green sensors compared to blue or red as usual? This should affect the color balance.
How many pixels will you use in the first place? If there is many each will be more noisy, but spacial resolution should be better.

All of these choices will lead to different original 1s and 0s, even before any post-processing.

[–] General_Effort@lemmy.world 3 points 9 months ago

Are the raw photos manipulated or are they just the original 1s and 0s unedited?

I think your average camera does not have the option to save RAW files. It seems somewhat common even outside professional equipment, though.

Could upload it’s hash

Yes, exactly. However, this would only prove that the image (and metadata like GPS coordinates) existed at that particular point in time. That would add a lot of credibility to, say, dashcam footage after a collision. It's curious that misinformation has become a major issue in the public consciousness, at a time when we have far better means of credibly documenting facts, than ever before.

But it would do little to add weight to images from, say, war zones. Knowing that a particular image or video existed at a particular point in time would rarely allow the conclusion if it was real or misinformation. In some cases, one may be able to cross-references with independent, trustworthy sources, like reporters from neutral countries or satellite imagery.

Creating a tamper-proof camera is a fool's errand. The best you can do is tamper-resistant. That may be enough if the camera can be checked by a trustworthy organization and does not leave its control for long. But in such a scenario you would rarely need that, and it's not the usual scenario. The price would be very high. Fakes that do pass muster will be given more credibility.