What is valid evidence in a world with perfect CGI technology?

Question

There are already real-world problems with analyzing whether particular graphic evidence is "photo-shopped" or genuine. In sci-fi technologically advanced future, realistic and perfect computer generated images are pretty much standard. Same applies to audio.

In past, graphic and audio manipulation was only for professionals. Now, simpler tasks such as removing pimples or altering (obfuscating) person's voice is something everyone can do, often with little effort. We can go off a hyperbole here and expect people to be able to assemble realistic photos of things that didn't happen at all. Details about how this is done are not point of this question, but it is perfect.

Under such circumstances, is there any reliable evidence? Or does that put world back in the dark ages, where you can only rely on people's testimony?

I've been thinking, but did not figure out anything that could be used as universal evidence that someone did or said something. Is there something I missed?

Answer 1

Cryptography.

If the surveillance camera contains a private key which it uses to sign its footage, that's a piece of evidence you cannot fake without having access to the camera's private key. Any piece of digital data could be required to have such a cryptographic signature in order to count as evidence.

To prove that some image or other digital data already existed at a certain time, its hash could be entered into a public database that keeps track of the time of entry (this could be a government database, or something like the Bitcoin blockchain).

Digital trace

Already today we generate a digital trace throughout the day, from location data of our mobile phone, to places where we used a credit card, to places where we happened to walk by some surveillance camera that we didn't even notice. In that future, the digital trail will probably contain much more detail. While you can fake single pieces of evidence, faking a complete consistent digital trail is a completely different issue. For example, say you manufactured evidence that you were in a specific shop at a specific time (and therefore could not have done the crime which happened elsewhere at that time), then the police would request the shop's customer tracking data. And then, if the customer tracking data doesn't show you in there, you'll have a hard time to explain. Oh, and that automatic update of your XY app, which happened to run right when the crime happened, came from an IP handed out by a wireless network that covers the crime scene.

Answer 2

I think you are missing the fact that generally speaking, even in courts of today, no single piece of evidence is sufficient for a guilty verdict.

Sure, technically speaking, you can make an image that is awfully difficult to tell that it is fake, and present that as evidence in a case. But if the other side disagrees, they will present their own evidence which contradicts yours.

If both sides present equal amounts of equally valid (in the eyes of the court) evidence, most modern societies fall back to the principle of presumption of innocense. If either side is able to present more or better evidence for their view, then that side usually wins. This is how courts already operate, and while it isn't perfect, it's pretty widely believed to be better than the alternatives.

So if, in your world, perfect faked images and audio clips are commonplace, then the courts would have to operate within the assumption that any presented image, video or audio may have been tampered with or outright faked. As a result, the evidence value of those will be reduced, possibly to the point of being a mere curiosity rather than considered valid evidence.

Suppose that in such a legal environment, I claim that you robbed me in the street, and I am able to present a photograph allegedly showing the event. The picture is good enough that everyone can readily see that you are holding a gun to my head while I drop my wallet on the ground, there are no signs that the image has been tampered with, and all of the image metadata corroborates my story. How might you defend against this accusation?

You might defend against my accusations by providing alternative evidence showing that you were somewhere else. Phone metadata exists in today's world, and very few people in the Western world do not regularly carry a cell phone with them. If you are able to show that the cell phone company had your phone moving through a different part of town at the time of the alleged crime, that:

• shows that you might not have been where I claim you were at the time (at least, it shows that your phone wasn't at the location where I claim you were)
• brings in a third party, which presumably has no particular interest in the outcome of the trial, and thus can be assumed to be reasonably impartial -- this is already the defining principle of witness testimony

Note that cell phone metadata is only one example; there are many others. Another example would be that a bank might be able to confirm that you made a purchase five minutes earlier on the other side of town -- sufficiently far away in terms of distance, and close enough in time, that you couldn't possibly have been where I claim you were at the time when I claim you were there. There is also the seemingly growing abundance of monitoring cameras throughout our society, which could also provide corroborating evidence from third parties. And of course, human witnesses (hey, I was walking my dog and stopped to chat with the neighbor around the alleged time; ask the neighbor, and check their phone location records to see if they agree with those for my phone!) If you are able to present multiple such pieces of evidence that support your story of you being innocent, and I am unable to properly address that discrepancy, at the very least that seems likely to lead to a not guilty verdict, and may also get myself into trouble as having given false testimony.

In the end, evidence which can easily be faked is trusted far less by courts than evidence which is non-trivial or very difficult to fake. Assuming that the justice system remains similar to ours, this would seem likely to not change in your world compared to ours; only the relative trustworthiness of different classes of evidence would change.

It also seems likely that, in a world where much evidence can be easily faked, the penalty for false testimony would be much harsher than it might be today.

Of course, a not guilty verdict does not necessarily imply innocense, but again, that's a trade-off where a society has to choose which alternative is worse. It's either living with the risk of people who committed crimes being pronounced not guilty and walking away, or living with the risk of people who did nothing wrong being pronounced guilty and being sentenced for a crime that they did not commit.

Answer 3

Secure storage / chain of custody

It is already the same in nowadays courts. Whenever you present some piece of evidence in court the first question will be "how have you obtained this piece of evidence?" If your evidence could have easily been faked you will either have to provide an expert which can assert your evidence is an original (not possible with perfect CG)

Or you have to present where you have obtained the evidence. If it is CCTV footage, someone has to have given you the tape - the security guard of the place owning the CCTV camera can testify that the tape you present is indeed the unaltered image which is also recorded in the secure temper-resistant storage of their secure server. They can even send a security-expert to verify that the secure storage is still safe and the original is the same as presented in court.

If you have an audio tape you will need a witness who recorded the tape and can testify that it is original. You can also build temper-resistant recording devices, which will record GPS, ambient noise and more together with audio to a write-once protected storage.

Answer 4

DNA, fingerprints, ballistic forensics and other types of "old-school" forensics would still be just as valid as they are now.

The police finds a naked woman and her lover shot dead in a bed. In the woman's husband's house, they find a gun. The gun is registered in the woman's husband's name and has his fingerprints all over it. The bullets from the crime scene are all perfect matches to the gun. Hairs matching the husband's DNA are found at the crime scene. The husband's fingerprints are found on doorknobs at the crime scene. Dirt from outside the crime scene is found underneath the husband's boots and the tires of his truck. Tracks exactly matching the tires of the husband's truck are found outside the crime scene. They show the vehicle aggressively skidding to a stop outside the building. Footprints made by his boots lead into the building, into the apartment and all the way to the bed. No video. No eyewitness testimony, even. Yet plenty of evidence to convict.

That was just an extreme example or an "orgy of evidence" where I listed some examples of "old school" types of evidence.

Witness duty

In an age where all photographs are assumed to be potentially fake, the police may develop certain procedures where they, after they have finished their crime scene investigation, bring dozens of eyewitnesses to the crime scene, observing all the evidence and making notes. These witnesses may be asked to appear in court as a substitute for photographs of the crime scene.

Answer 5

I'm not a technical expert but then I suspect I don't need to be because your question states that the 'evidence' can be fabricated to a point where it's simply not possible to prove it isn't real. So with that in mind I think what you end up with is a shift in how that evidence is used.

This sort of evidence (lets say cc tv of a person performing a street robbery) is given a lot of weight because at the moment it's much more likely to be real than fabricated. Especially if the person on trial is a petty criminal or has no previous history of crime. Also the theft of a small amount of money would suggest that the expenditure of time and effort to fake it isn't reasonable. If it was cc tv of a world leader being assassinated then it may well be in the interests of an organisation to make a fake.

When there is no way to know if that cc tv is real or not then it becomes no more useful than the testimony of a third party saying that is what they saw. In fact it's probably slightly less useful as the person can be questioned where as the piece of tape can't.

I think perhaps that it would still be admissible but given no more weight in legal terms than circumstantial or unverifiable evidence is today. Just like the fictional 'gangster' can provide a number of eye witness accounts that he was out of town the night of the shooting by either buying or relying on the loyalty of his mob, he can now also provide digital 'proof' as well.

If you allow your high tech to be able to fabricate forensic evidence as well as visual and audio then it's even worse and we are down to simply interviewing people on the witness stand and making up our minds based on story inconsistencies and overall reliability of their testimony.

Answer 6

One method I've seen to approach this is the development of one-shot-record-multi-read-only devices made as indestructible as possible. Attempting to get into or influencing the device to alter it results in the destruction of the recording.

While it cannot prevent loss of the recording, which has its own implications, it can prevent tapering.

However, it may not be able to prevent deliberately staged events unless there was also some sort of precise location tracking that either was completely internal or is external and somehow unblockable and unspoofable in and of itself.

As mentioned in the comments: tamper-evident is almost always as good as tamper-proof for evidentiary purposes.

Answer 7

In addition to the cryptography, digital trace, and secure storage/chain of custody answers already provided, also consider the possibility of statiscial analysis that can identify manipulation in subtle ways.

Here is a link to an article that describes some of these analyses:

https://articles.forensicfocus.com/2013/08/22/detecting-forged-altered-images/

This article discusses how simply opening and saving a file can change the structure of a picture without producing visual differences; how inserting a new image into another picture will leave mathematical artifacts that can be discovered with error analysis; and how subtle differences in quality in different portions of the same photo can be proof that the image was manipulated.

Now, there may be techniques that can be used that might mitigate some of this: one idea that comes to mind is to perfectly model a scene, and then create a perfect CGI with no artifacts. Even in a case like this, however, it may be possible to show statistically that the scene is fake: digital algorithms themselves often leave certain statistical "markings" that might be identifiable -- and, for that matter, going back to the quality issue, such a generated scene might have a quality that doesn't match the available video equipment that allegedly filmed the incident (ie, grainy when the camera would have had a high amount of detail, or highly-detailed when the camera would have been grainy).

Finally, it would be important to corroborate video evidence with details from the real world. A forger, for example, may have known that the individual in question wore a popular designer jacket, but would not have been aware of the tear in the jacket that happened at a party the night before, in a spectacularly embarrassing incident that everyone saw -- or, alternatively, that photo that the prosecution is claiming to be fake also happens to show a name-tag sewn on the jacket, that the person alleged to have faked the photo would have no reason to know actually existed...

Answer 8

To prevent pictures/videos/sound files from being edited afterwards, use Trusted timestamping:

CC-BY-SA-2.0 Tsuruya

With that technology, you can easily and reliably check when the data has been timestamped. The person was killed on Tuesday, the CCTV shot was timestamped on Wednesday: Something is not right.

Data without a trusted timestamped would not be considered valid. If you tamper with the data and want to use it as a proof, you have to timestamp it, but the date would be at least an hour late, or you would have to bribe the timestamping authority, which we can assume is not realistic for most criminals. The timestamping authority could be a well-protected autonomous service within the United Nations, so that it is resilient to even governments pressure.

The only strategy for the criminal would be to perform all of these steps within a few seconds, probably assisted by some artificial intelligence:

• Take the picture/video/sound
• Decide what to remove, what to add
• Perform the transformation with perfect CGI (presumably much more resource-intensive than your usual face-swapping app)

This would limit abuse to only very carefully planned crimes. Once timestamped, there is no way to go back to fix any inconsistency.

As pointed out by Peter Cordes, trusted timestamping does not protect against pictures/files that have been photoshopped in advance, before the actual crime took place. In this case, you have to predict everything in advance: the state of the trash bin, the shape of the blood stains, the sun and shadows, etc.

Answer 9

Lot of answers already but it's an interesting question, so my two cents worth:

This isn't a new issue. I read an article in a computer magazine decades ago where the writer argued that with the growing power of image-editing software, photographic evidence would soon be almost useless. Video is a lot harder than a single still, but as you say, we're getting there. Anyway ...

First, I'd question the premise. Could any CGI really be perfect? Yes, the technology to produce better CGI is advancing. But so is the technology to analyze images. I can readily imagine police experts (or whatever experts) carefully examining a video and saying, Wait, this is clearly faked, because on frame 6892 the shadow cast by the person in the video is 1/2 of an inch from where it would really be under these lighting conditions. Or, after running up the stairs like that he would surely be breathing heavily, but his chest is not rising and falling any faster than it was in earlier frames. Etc.

Second, even if it's perfect in that sense, an analyst might be able to demonstrate that aspects of the image are exactly what you get when you use Fwacbar CGI Creator version 8.2. A few years ago a reporter claimed to have uncovered embarrassing memos about the military career of a certain presidential candidate. He even had the memos posted on his network's web site. And he was quickly shot down, because the memos were supposed to have been written in 1973 but were EXACTLY what you would get by typing that same text into modern versions of Microsoft Word using default fonts, margins, etc. The reporter tried to argue that someone in the 1970s COULD have produced such a document using printing equipment available at the time, but many asked, Which is more likely? That someone writing a routine memo would use expensive and advanced printing equipment instead of an ordinary typewriter, and would just happen to use exactly the right combination of equipment to reproduce what Microsoft Word would do by default 30 years later, or that the documents are forgeries created using Microsoft Word a few days or weeks before they were "discovered"?

Third, almost all evidence used by courts today could be faked. A witness can lie. The murder weapon could be planted in the suspect's house. Documents can be forged. DNA samples could be planted. Expert witnesses can exaggerate the significance of forensic evidence. If the police or someone working with law enforcement are trying to railroad someone, lab results can be falsified. Etc. With sufficient skill and effort devoted to the task, I'd guess any evidence could be faked; I don't know what would be 100% reliable. That's why the standard of evidence in a criminal case in the U.S. is not "absolute, 100% proof" but "proof beyond a reasonable doubt".

Just as courts today are well aware that, say, a signature might be forged, so they'd recognize that a video might be forged.

They'd look at the source of the video. A video showing the defendant far from the scene of the crime would be much less persuasive if it was produced by the defendant himself than if it was from, say, a store's security camera where the store has no connection to the crime other than serving as an alibi. Does the store owner have any reason to create a fake video to help the defendant? If not, strong argument against it being fake.

They'd look for corroborating evidence and examine consistency of evidence. If there's a video AND there are fingerprints AND the murder weapon was found in the victim's car AND etc, that's much more persuasive than if the only evidence is the video.

Side thought: Several other posters talk about technical means of authenticating videos, time stamps and encryption keys and so forth. Sure, any technical means of authentication would help, but presumably any of that could be forged by a sufficiently skilled hacker. If you're producing the fake video from scratch yourself, presumably you could put on all the authentication et al AFTER you produce the video. If the premise is that we're assuming that the technology exists to produce a 100% perfect, undetectable CGI, it seems to me that we're violating the premise if we say "but ... there's no way to beat an authentication key".

Answer 10

Will fakes ever be as good as real? While technology for faking evidence gets better, so does the technology for gathering evidence. In Iain M. Bank's Player of Games such a world exists and they say a a drone can provide a such a recording but a ship mind is required to be in contact. That would come down of course to trusting the gatherer of the evidence.

Thinking about it, let's say you have a source of entangled photons in a safe location. One of each pair goes into a fiber-optic cable to a camera and used to light the scene and the other is protectively stored in the bank. The "Camera" records the actual photons from the scene along with a hash of the data in each frame and transmits the hash back. To verify the video you could validate that the video images were taken at that time and in the proper sequence by looking at their entangled pairs. Maybe viewing the data requires both entangled pairs and it can only be viewed once, though at the time of extraction there could be a recording made of its viewing and verification that could be attested to.