News for hackers. Stuff that matters.
Re:NeRF: An eventual successor for deepfakes? by irjustin on Monday June 27, 2022 @10:49AM (#31895339)
This has been discussed many times and it doesn't work.

Simple answer is I can just record a deep fake and get it cryptographically signed.

Re:NeRF: An eventual successor for deepfakes? by andruby on Monday June 27, 2022 @10:59AM (#31895491)
If you trust a person (or source) and they have a private key that they can properly secure, they could always sign their material with that key. That would prove that the source provided that material.

A blockchain could be a way to store and publish that signature & hash.

It can't say "this is real", it can only say "that signature belongs to source X".

Re:NeRF: An eventual successor for deepfakes? by blamestross on Monday June 27, 2022 @11:19AM (#31895781)
> A blockchain could be a way to store and publish that signature & hash.

yes but it would be a bad one. We have multiple key distribution mechanism that are better for the use case.

Re:NeRF: An eventual successor for deepfakes? by intrasight on Monday June 27, 2022 @01:04PM (#31897442)
I disagree. The key alone is not sufficient nor secure. We will need crowdsourced validity data as well. We need a zero-trust model - and I too believe that blockchains will play a role.
Re:NeRF: An eventual successor for deepfakes? by notfed on Monday June 27, 2022 @01:41PM (#31897974)
If a video is already cryptographcally signed, then you can safely distribute the signature on an untrusted channel.

Adding Blockchain into the mix is superfluous, and destroys scalability.

Re:NeRF: An eventual successor for deepfakes? by intrasight on Monday June 27, 2022 @01:58PM (#31898205)
We don't watch signature keys - we watch videos.

The TV will have to match every short segment - perhaps 5 seconds of video - against a blockchain which scores the validity of that segment - and of course looking back to it's original source. Signing the whole video is necessary but not sufficient.

But yes, this is going to be resource-intensive.

Re:NeRF: An eventual successor for deepfakes? by blamestross on Monday June 27, 2022 @02:44PM (#31898869)
> against a blockchain which scores the validity of that segment

Why not just allow a cert with that information to be delivered alongside the video? Where would the "score" come from?

Re:NeRF: An eventual successor for deepfakes? by intrasight on Tuesday June 28, 2022 @11:14AM (#31908796)
score is ai and/or crowdsourced and comes from a special blockchain
Re:NeRF: An eventual successor for deepfakes? by blamestross on Tuesday June 28, 2022 @01:26PM (#31910496)
Why would either of those things need a block-chain?

Crowd-sourced is already "network of trust".

An AI based score would have to be produced by a centralized provider, so network of trust is the reality for that too.

The only way blockchains would provide benefit would be as a distributed discovery mechanism for "review" of the video chunk and having an open ecosystem for that (a dht or trackers) would work better.

Blockchains only ever had a reasonable use case under the assumptions of functional capitalism (and we don't have one of those). The reality is that they can't be sustainable without capture and the market incentives only increase the incentive to capture it.

DHTs and Networks of Trust only have the value of what service they provide and while that is less exciting for scamming people, they have survived and been high functioning for decades.

Re:NeRF: An eventual successor for deepfakes? by blamestross on Monday June 27, 2022 @01:07PM (#31897486)
Zero trust models don't exist and the laws of physics (probably) don't provide for them. (Materialism is a real problem in physics nowadays)