Given the resources available to
A) a national news agency like the NYT or NBC
or
B) the NSA or other American government agency
how hard would it be to definitively declare that an audio or video tape was authentic or phony if it were made by
Money.
Time.
The more money and the more time you have, the easier it gets to produce a more convincing fake.
A perfect fake? I don’t know. A fake good enough to convince someone? It depends how much time and money they have, to carry out their investigation.
I’m pretty good (not great, but pretty good) at editing audio recordings. Most people can’t tell where my edits are. But even with my very amateur program, I can actually see them.
Imagine that speech is like a perfect sine wave. Every place you add or chop something, there are minor differences in background noise, the speaker’s inflection, the natural rhythm of words, the length of time the speaker pauses to take a breath. Suddenly, the waves are spaced differently, the highs and lows change, and so on.
Given enough time to play the recording over and over, and equipment good enough to capture those minor variations, a good technician could spot the edits.
But if I understand you correctly, you’re talking about the “original” fake, if that makes sense. Couldn’t the fake be run through some kind of program that looks for exactly what you’re talking about, and digitally adds or subtracts whatever is needed to make all the transitions indistinguishable from normal variations in background noise, breathing, etc.?
Exactly… And how good the equipment needs to be (i.e. money), and how long it will take to get a final result, depend on the time and money put into the fake.
If it’s fake.
Proving a fake is, at least in theory, possible. Proving a non-fake - proving the negative, that a piece of media has not been tampered with - is theoretically impossible. You can be pretty sure, and no more than that. Theoretically.
Given the rapid pace of technology development, it’s undoubtedly going to get very much harder very quickly.
It also depends on just what is being claimed of the recording. It’s a lot easier to get away with “this is a recording of a senior KGB official, made some time in 2017” than it is with “this is a recording of Vladmir Putin, from November 12, 2017”. In the latter case, you’d have to match Putin’s voiceprint perfectly, and the recording could be debunked by other evidence is Putin’s entire daily schedule for that day were known. In the former case, you just need any old voice actor in any old office, and who’s to say that that voice actor isn’t a KGB agent, anyway?
Every voice pattern is different. And every “programmer” would have their own particular quirks. And that’s just audio. Now add video. Facial tics, tilting of the head, the amount of light from the window in the background. . .
Let’s put it this way. You know how some people watch the CGI in films just so they can point out how flawed it is? Now turn those nitpickers loose on a pornographic video of a world leader.