What will be our defense when CGI is indistinguishable from live recordings?

I’ve pondered how to start this thread without rambling for quite a while now and have decided to just bite the bullet.

Currently, having a video of someone doing something is considered almost absolutely reliable as evidence. I see this changing sooner rather than later.

Computer graphics are getting better ad better at an almost exponential rate. Eventually some public figure is going to be accused of some crime or malfeasance based on a video that will ultimately be proven to be faked.

Take a look at the trailer for the upcoming game Wolfenstein II: The New Colossus. Full CGI starts around 1:45.

Yes, this is still obviously computer generated even without the fantasy components. But also consider the details. Seamless clothing and facial features. Realistic facial expressions with synchronized speech.

Some of the trailer is rendered in real time on a high end home gaming computer. What if instead of HD quality it was being rendered at typical security camera resolution and focus. I’d bet even a skilled tech would be hard pressed to persuade a jury that a frame-up video was faked.

In a decade or two, how will we prove the video of the president kicking puppies didn’t happen?

Or flipping it around, there’s a real video of the governor taking a bribe. But then his defense team shows up with CGI work files (wire frame models, texture maps, CGI rendering of the room) claiming that it only ever happened on someone’s computer.

Off the top of my head I would say the same standard as before when it was more just still images; multiple angles from unrelated sources and/or eyewitness testimony.

When it becomes impossible to distinguish between real and fake video footage then video footage will no longer be credible evidence.

Our legal system functioned before video evidence existed and it will continue to function after video evidence passes into irrelevance.

Aren’t we kinda there? Ellen Degeneres has/had a show and I remember her crew making videos which made it look like Obama was doing funny stuff. I can’t find a link at the moment but it looked real.

As for the video linked in the OP, it might look pretty but it’s not anywhere near photorealistic. The kind of fake videos which would matter most would tend to be videos where a human is featured and getting photorealistic faces, especially with proper subsurface scattering and animations, is rather far away and perhaps impossible.

If that freaks you out. . .Radiolab recently did a show about video manipulation. It’s hard to explain exactly what is done, and the end product is still obviously manipulated, but it’s just a matter of time til when it wont be so obvious.

“Breaking News”
Future of Fake News


As implied in the OP, I think we’re basically there now. People take photoshopped stills as real and the only reason we don’t see fake video more often is it’s slow to produce, not that the quality is not there.

The reason we don’t see people get framed very often is the same as why framing generally isn’t more common; the person you are framing is not a willing accomplice. They will be protesting their innocence and saying the video is faked. If then any material evidence supports their version of events then perhaps people do start looking again at this video and who made it.

Sadly I think the risk reward might make sense in politics though. Get some video out there that makes a politician look bad; you don’t need them to be charged of any crime.
But hey, we already knew democracy is broken.

Another subject that I know something about (and get paid very well for knowing).

First, CGI has a VERY long way to go to reach the point where it is even close to “real” video.

Second, video forensics is very similar to document forensics. One might as well say that photocopying a page produces a document that is indistinguishable from the original. To an expert, there are obvious differences.

On the other hand, I have to admit that any video or image is simply a set of pixels with certain numerical values. In theory, it is possible to manipulate the values of pixels to make a video or image that is absolutely identical to a video or image created in “real time” using a camera. But I’ve never seen one that came close.

We’ve also reached the point on the opposite end of the spectrum; people refusing to believe genuine evidence because they argue it’s been faked. Look at all the people who refused to accept the evidence of Obama’s birth certificate even after it was made public.

Yes, and there’s a big difference between a viral video that shows the President kicking puppies and a surveillance video submitted as evidence that a suspect used a firearm in a robbery.

In the first instance, people will believe what they want to believe. Hopefully, they will be aware of what can be done with editing and rendering.

In the second case, there’s a much higher bar to overcome when a prosecutor (or defense attorney) produces a video or image as legal evidence for a trial. Chain of custody, provenance, supporting evidence, etc., become big issues. A prosecutor doesn’t just walk in like they do in TV shows and say, “Here’s a video obtained at the last minute from a mysterious source that PROVES this guy used a gun!”

One possible solution is that manufacturers of video equipment will include cryptographic signing.

It is possible to create a video file that includes information like “this was shot in <location> at <time> by <hardware>” in a way that is very difficult to fake, and that won’t persist through a modification.

Educating people about only believing video that’s been verified is going to be, well, difficult. We can’t even convince people not to believe the bullshit their friends share on facebook.

But for people/organizations who want to verify that video is legit (say, reputable news organizations and courts of law), untampered with, and unedited, there are technological solutions.

In the law, the term is “authenticating” a document or video. That is, you generally need to have someone testify to its authenticity before an item of evidence can be admitted. Only certain documents (generally, government records and other things that are traditionally reliable) are self-authenticating.

So, for example, if you wish to admit into evidence a picture of a crime scene, you usually need testimony from the photographer testifying that they took the picture. At a minimum, you’ll need somebody to testify to its reliability, based on their personal knowledge that it is an accurate representation of the actual scene.

So, even now, the law imposes a standard of verification on pictures and videos.

I like the cryptographic signing, but it’s probably going to take some government regulations to get everyone on board with that and we now how well that will go over.

Moriarity. While courts of law have standards, the court of public opinion is a whole other genre of animal. And that court is where politicians and other celebrities are made and broken.

No, it won’t. All it will take is technical mechanisms and it will serve its purpose: The legal standard won’t change, it will just have more evidence to work with, including how reliable the firmware of various cameras is known to be.

It’s like using a receipt from a store to prove you were in a specific location at a specific time: Does the receipt look forged? Is the store known to have a bad clock in its POS terminal? Digitally signing an image or video file with a datestamp and other information is technically more involved, but it comes down to the same kinds of questions. There’s no need to legislate any of this, especially given that there’s a built-in market for it among journalists at the very least.

I don’t know about video, but most still images are already digitally signed/date marked. All you need is a free program (if your computer doesn’t already have one) that can pull the EXIF data. It’ll typically give you the date and time that the picture was taken (or at least what the camera thinks the time was), the camera settings (focal length, shutter speed, aperture, etc), model (and SN?) of camera, in some cases GPS data can be stored in there as well and probably a bunch of other stuff that I haven’t kept up with since I stopped taking pictures with a ‘real’ camera a while back.

However, it’s also trivially easily to strip the EXIF data from the picture but I’d guess that the majority of people don’t even know it’s there.

Well, I mean, it’s been a while since a well photoshopped still picture and a real still picture have been indistinguishable. We manage to overcome this hurdle by, for instance, trusting the source’s reliability: if you see a random picture in some internet board, you don’t trust it, if it’s in some respected news site, you do.

Getting all the manufacturers to agree on one standard of encryption won’t be easy without some sort of incentive. Most of them won’t even be willing to put forth the effort because there’s nothing really in it for them. Or if there is something to gain from it they’ll try to implement it in some way to exclude or at least severely inconvenience everyone else. Just look to Microsoft and Apple for examples of closed architecture and proprietary standards. Smartphone manufacturers won’t even agree to a common interface. The only universal incentive for them is a big regulatory stick. I’m not a big fan of government regulation, but I do recognize that sometimes it is needed.

Certainly real vs edited video can be authenticated through current digital forensic techniques. However there is a difference between defending a piece of digital evidence in a court of law and defending a piece of video in the court of public opinion. People are already swayed by “fake news” and crude photoshoped images. And you don’t have to convince everyone. Just enough people to make a lot of noise.

No, not cryptographically, which is what “digitally signed” means in this context. Cryptographic digital signatures make it computationally infeasible to alter an image unless you’ve extracted a secret key from the camera.

It’s happened in the past. All web browsers support the same image formats, for example.

True enough, but my point was just to note that the legal system, long ago, recognized that some piece of evidence may not be what it appears to be (again, with exceptions, based on their reliability. This creates an interesting list, as, for example, family bibles are generally self-authenticating for determining genealogy), and that its provenance should be verified by some independent witness.

Perhaps the court of public opinion should adopt the same skepticism. Some picture or video should not be presumed authentic unless you know who shot it and have had a chance to appraise that person’s credibility.

I imagine that this is the sort of tactic that will gradually become semi-discredited after a few instances of hoaxes or framing-an-innocent-person have happened. Eventually the public would (hopefully) treat video evidence with a bit more suspicion.

This could, of course, also allow genuinely people to get off scot-free (“That video is CGI!”) but that just comes with the territory.