A Site for Photographers by Photographers

Community > Forums > Digital Darkroom > Imaging Techniques>Other > How to tell if a photo has...

Featured Equipment Deals

Craft Photography Fundamentals Read More

Craft Photography Fundamentals

Simple tips for beautiful blog photography with a focus on crafts by A Bit of Sunshine's Rebekah Gough

Latest Equipment Articles

Lensbaby Spark Review Read More

Lensbaby Spark Review

This inexpensive gadget does indeed spark your creativity. Read on to see how.

Latest Learning Articles

26 Creative Photos of Water Drops Read More

26 Creative Photos of Water Drops

These absolutely amazing macro photographs feature a tiny elemental thing that can hold a lot of mystery. Take a moment to enjoy these photographs of water drops.


How to tell if a photo has been edited?

Ross Thompson , Mar 20, 2004; 06:37 a.m.

Is there any way of tellig or any software that will tell me if a photo has been edited? Furthermore, is there any software that will restore it to how it orignallly looked? I was wondering whether something and someone had been edited out of some photos I have been sent and wondered if there was any way of telling for sure?

If anyone knows how or knows any programs that can do it, I would really appreciate your help. Cheers Ross

Responses


    1   |   2     Next    Last

Leonard Evens , Mar 20, 2004; 10:03 a.m.

Response to How do I...

I am fairly confident that it is possible, using the current public/private key cryptography methodology, to introduce additional information in an image which would allow anyone to detect a forgery and even to recover the original image. Certainly a digital signature can be added to an image in encoded form which any manipulation would destroy. If the signer tells us how, such a signature can be read by anyone but not modified without destroying the signature and possibly the image. I don't believe however that many images produced by present day technology incorporate such features and I don't think there are as yet any standard ways to do it.

Now and in the immediate future, I think you have to examine an image carefully. A lot of digital editing, even that meant to mislead, is done clumsily and careful examination of the image will demonstrate it. A good example is the widely circulated photo in which images were combined to show a young John Kerry standing next to Jane Fonda. It seems clear to me from the perspective and subtleties in the lighting that the image is a forgery.

David Loose , Mar 20, 2004; 10:52 a.m.

Response to How do I...

The Canon 1DS is (advertised as) the only camera that can record a digital signature that can be verified after the fact. That is only useful if someone passes the image along with the signature, it's certainly possible to edit the image and save it without a signature.

Automatic restoration is definitely not possible, unless someone sends you a Photoshop file with layers where you can see the modifications and remove them, which doesn't seem likely.

Unless that's the case, there's no software that can tell you an image is not edited... You can look carefully, but it's your word against the picture.

Mark Herring , Mar 20, 2004; 01:43 p.m.

Response to How do I...

For normal stuff--eg a photoshop file or a jpg, there is no way to know what editing was done, nor is there any way to reverse it.

Depending on the quality of the editing, you may be able to see that something was done, but that's about it.

In reply to Leonard's answer: There's no question that the technology is available to detect if an image is altered. But to include information that would allow restoration of the original is a very different problem.

David Loose , Mar 20, 2004; 02:36 p.m.

Can you post the picture? I imagine we'd all offer our expert opinions...

Brian Mottershead , Mar 20, 2004; 03:59 p.m.

I don't see how you could determine whether an image has been modified, in the general case. For example, load an unmodified image into Photoshop. Pick one of the 3 million pixels at random and add 1 to its RGB value. It is now a "modified" image, but there is no way that the eye could see the change, or that a software could tell that the value is not the original one. How would software know what the "correct" pixel value was supposed to be.

That one pixel change is far less change in RGB values than a color space conversion would be, which is not considered a "modification". Other global changes such as contrast and brightness adjustments are also not considered "modifications" and produce more global changes than the one I mentioned.

One could certainly generate a digest code for an image that would allow you to determine whether it had been changed since the digest was produced, just as you can with any digital data. Of course any transformation of the image from raw file that was used to generate the digest would show up as a modification, even one not normally thought of as a "modification", such as the color space conversion that I mentioned, and some of those transformations are not reversible since they may lose data in some parts of the image.

Paul Watson , Mar 20, 2004; 05:15 p.m.

1. Any cypher can be broken and tampered with. So there is no foolproof method of maintaining an images integrity. Remember our wonderful digital photographs are just digital files like any other. Crypthographers will come out with a new method and promptly be hacked. It is a leap frog race as usual.

Otherwise you should read this Wired article. It is about fake celebrity porn but don't let that put you off. It discusses how good some fakers can be in editing images that detectives have a hard time telling whether a shot is real or not. http://www.wired.com/wired/archive/11.11/fakers.html

2. There is no versioning in any image format I know of once saved. So if someone crops half the photo off you are not going to be able to "uncrop" it later on unless you have their Photoshop/Fireworks/PSP/etc. undo data.

David Loose , Mar 20, 2004; 06:02 p.m.

I agree in principle with what's been stated above, but I think it's a little overboard to say that any cipher can be hacked. I don't think ciphers like Triple DES, AES, and PGP have been hacked, using lengthy keys (128 bits and up).

There are plenty of examples in the news where people develop (poorly) ciphers that are readily cracked, but don't think it's fair to state that that's the case in general.

Brian, while it's true that color space conversion, brightness etc aren't considered "modifications" (especially in photo.net terms), those type of modifications are (normally) unnecessary when using a photo as evidence in court, which is the purpose of Canon's Data Verification sw. Another recent thread discusses the need for proving the chain of evidence, I think having it start with a digitally signed photograph is a good point.

Steven Clark , Mar 20, 2004; 06:51 p.m.

In theory, without a copy of the private key no-one can generate a new false signiture. And by the definition of how a good one-way hash algorithm works it should be essentially impossible (possible but a computer problem that requires such calculation it makes cracking a 1024-bit key look like 1+1=2) to given an existing signiture find ANY other image that fits that signature, and approaching 0 probability that if you did it would be the image you wanted. I'd call that a pretty secure algorithm. Also, since you would be encrypting a much smaller hash digest of the image rather than the image itself for the signature you can afford to use really rediculously huge keys that would otherwise be too slow to consider so that 20 years from now it will still take the entire worlds computing power to crack the key before the statute of limitations expires. I'd call that a reasonably secure system wouldn't you?

The weak point of course would be if someone managed to get the private key off the camera. The issue here would be designing the camera such that it would be impossible to extract the key without destroying it or the camera itself. For example embedding it along with some basic encryption into a seperate chip so that it can be used, but not accessed by the camera itself would mean that you'd physically need to somehow tear apart and examine the chip. You'd still of course need to trust Canon not to do something stupid like archive copies of the private keys, or allow someone on the floor enough access to manufacture a camera with a known key, but at that point you're starting to deal with conspiracy theory.

Gordon Richardson , Mar 21, 2004; 12:47 a.m.

New Scientist magazine had an article a few months ago about statistical techniques (based on edge detection) to find manipulation. (I can't find the online article, since you have to register to search...) The consensus seemed to be that you can catch the really clumsy cheats, but not the sophisticated ones. So, you probably can't really tell...


    1   |   2     Next    Last

Back to top

Notify me of Responses