Skip to content
Photo evidence7 min read

How to Prove a Photo Hasn't Been Edited

To prove a photo has not been edited, seal it at capture: compute a fingerprint of the file, bind it to a date certified by a third party, and sign it. Any later change is then detectable by anyone.

By Cédric Driaux, Founder of tryVidimus

Drawing of a photo with a crop frame and a row of coloured blocks

Why "it looks real" isn't proof

Most people try to prove a photo is not edited by pointing at it: look, it is obviously real. That argument no longer works. Editing tools are on every phone, and a careful edit leaves nothing the eye can catch.

Analysis software does not close the question either. It can flag suspicious areas, but it gives an opinion, with a margin of error. The other side can always answer that the tool is wrong.

So the useful question is not "does this photo look untouched?". It is "can someone who does not trust me check that this file is the one that was captured?". That check is one of the four conditions described in our photo evidence guide.

Consider a camera shop that rents out lenses. A lens comes back with a scratch on the front element. The shop has a photo of the clean lens taken at handover, and the customer replies that the photo could have been retouched.

The shop cannot prove a negative by looking harder at the picture.

Hashes explained simply

A hash is a fingerprint of a file. A public algorithm reads every byte of the file and produces a short string of characters. The one used most for this purpose is SHA-256.

Four properties make it useful to prove a photo has not been edited:

  • The same file always gives the same fingerprint, on any computer and with any tool that implements the algorithm.
  • The smallest change gives a completely different fingerprint. Changing one pixel, or one character of metadata, is enough.
  • The fingerprint does not reveal the photo. It can be shared or sent to a third party without disclosing the image.
  • No practical way is known to build a different file that has the same SHA-256 fingerprint.

The check is simple. Compute the fingerprint of the file you hold today and compare it with the one recorded at capture. If they match, the file is byte for byte the same.

Digital seals and C2PA

A seal ties the fingerprint to something the photo's owner cannot rewrite. Two mechanisms do this, and they complement each other.

The first is a timestamp. A timestamp authority receives the fingerprint, never the photos, and returns a signed token that binds this fingerprint to a date, following RFC 3161. From then on, a different file cannot be passed off as the one that existed at that date.

A companion article explains how an independent timestamp proves when a photo was taken.

The second is a signature carried by the photo itself. C2PA is the open standard behind Content Credentials. A signed manifest travels with the file, records its origin and makes any modification visible.

Our article What is C2PA? explains the standard in plain terms.

MechanismWhere it livesWhat it shows
SHA-256 fingerprintIn the proof recordThis file is byte for byte the one that was captured.
Qualified electronic timestampIn a signed token from the timestamp authorityThis fingerprint existed at a certified date.
C2PA signatureInside the photo fileWhere the photo comes from, and that it was not altered after signing.

tryVidimus applies the three together. Each file gets a SHA-256 fingerprint, each proof receives one qualified electronic timestamp, and each photo is signed with C2PA. The product is a C2PA conformant generator product for JPEG photos, at assurance level 1.

A verification example, step by step

Back to the camera shop, this time with sealed photos. At handover, the employee opens a capture link and photographs the lens live. The photo cannot come from the gallery, and the proof is then sealed.

A few weeks later the customer contests the scratch. The shop sends the link to the proof's public verification page. The customer, or their insurer, can then check the proof without an account and without asking the shop for anything.

  1. Check the file. Compute the SHA-256 fingerprint of the photo with any standard tool and compare it with the fingerprint in the proof.
  2. Check the date. The timestamp token is a standard RFC 3161 token, so it can be verified with OpenSSL, independently of us.
  3. Check the origin. Any tool that reads Content Credentials can open the C2PA manifest carried by the photo and confirm that the signature is intact.
  4. Check without our servers. The verifier is open source, under the MIT licence, so the whole check can be rerun on your own machine.

If someone had edited the photo after capture, the first step would fail: the fingerprint would no longer match. The documentation describes what a proof contains and how to verify it.

Notice what changed. The shop no longer asks to be believed. It hands over a check that gives the same result for everyone.

Limits: what a seal does not show

A seal is precise about what it covers, and it is worth being just as precise about what it leaves out.

  • It covers the file from the moment of sealing. It says nothing about what happened before, which is why capture has to be live, with no upload from the gallery.
  • It does not say the scene is truthful. A sealed photo of a staged scene is still a photo of a staged scene.
  • It does not say who held the phone, or what state the object was really in outside the frame.
  • It applies to the original file. A copy that was resized, compressed by a messaging app or captured as a screenshot is another file, with another fingerprint.
  • Metadata is not a seal. The date and device stored in the file can be rewritten, as explained in why EXIF metadata can be faked.

These limits follow from one principle: certify, never arbitrate. The seal settles the question of the file. The question of what the photo means stays with the parties, an insurer or a judge.

Frequently asked questions

Can you tell whether a photo has been edited just by looking at it?

Not reliably. A careful edit leaves nothing visible, and analysis tools give an opinion with a margin of error, which the other side can contest. The dependable approach works the other way round: seal the photo at capture, so that any later change breaks the fingerprint and the signature. The check then gives the same result for everyone.

What is a SHA-256 hash of a photo?

It is a fingerprint of the file: a short string of characters computed from every byte of the photo by a public algorithm. The same file always gives the same fingerprint, and the smallest change gives a different one. Comparing the fingerprint recorded at capture with the one computed today shows whether the file is still identical.

Does C2PA prove that a photo is real?

No. C2PA records where a file comes from and makes later modifications visible, through a signed manifest carried by the photo. It establishes the origin and the integrity of the file. It does not say that the scene in front of the camera was truthful, and it does not replace the judgment of whoever reads the photo.

Can I seal a photo I took last week?

Not with tryVidimus: capture is live only, with no upload from the gallery. A seal placed on an old photo would only prove that the file existed when it was sealed, not that it was untouched before. To prove a photo has not been edited, the seal has to be applied when the photo is taken.

What happens to the seal if I send the photo through a messaging app?

The seal stays valid for the original file, but the copy that arrives may no longer match it. Messaging apps usually compress images, which produces a different file with a different fingerprint. Share the link to the verification page instead of forwarding the picture, so the recipient checks the original proof.

Does the timestamp authority see my photos?

No. The timestamp authority receives the fingerprint of the proof, never the photos. A fingerprint does not allow anyone to rebuild the image. The authority returns a signed token that binds this fingerprint to a certified date, and that token can be verified later without contacting us.

Try it in the sandbox

You cannot prove a photo has not been edited by arguing that it looks real. You prove it with a fingerprint fixed at capture, a certified date and a signature that anyone can check. You can try the full journey for free in the sandbox, without a card.

Create an account

Official sources

Cédric Driaux · Founder of tryVidimus

Cédric Driaux is the founder of tryVidimus, published by Fyher. He has been a software developer for 23 years. He also builds ScanStay, a photo proof tool for short-term rental hosts.

Related articles