Launch HN: Reality Defender (YC W22) – Deepfake Detection Platform • Hacker News Board

Re: Launch HN: Reality Defender (YC W22) – Deepfake Detection Platform

#91

Post by btbuildem » Thu, Mar 24, 2022, 1:50 PM UTC

Geee wrote: post #29

We have solved this problem already for text with digital signatures. There might be a way to digitally sign speech in real-time and show the signature on a display device, which is captured on video. It could be just a smartphone app or a separate device. This way every video could be proven authentic, even if they originate from unofficial sources.

How do you hash something so fuzzy as video / speech? Two different microphones picking up the same audio will result in two digital recordings that hash differently.

Re: Launch HN: Reality Defender (YC W22) – Deepfake Detection Platform

#92

Post by asail77 » Fri, Mar 25, 2022, 1:12 AM UTC

miunau wrote: post #79

smt88 wrote: post #51

Earlier quoted context omitted.

You need cryptography for hashing and verifying the integrity of any copy of the database. You probably don't need proof of work, you're right.

Proof of work is not needed, just a pool of time servers that you can make a quorum out of. We're working on exactly this on top of roughtime ( https://roughtime.googlesource.com/roughtime ) for a future version of QR Date ( https://qrdate.org/ )

Great discussions! Some thoughts:

First, there are multiple solutions to this problem, all need to be explored and many will have a part in the future. Bad actors will do everything they can to find a way around every solution.

Second, for the hashing approach, have a look at "Perceptual Hashing", it's a way to hash content like a an image, even if the resolution changes.

Where to store the hashes? A centralized server is probably fine, but there is always a risk of a bad actor exposing it somehow. A secure blockchain can work better. But if you go that route, might as well go with the most secure blockchain. POW is generally more secure than POS. And currently the most secure blockchain is Bitcoin. So one solution is to batch hashes together and write them with a Bitcoin transaction in some cadence.

-Ali

Re: Launch HN: Reality Defender (YC W22) – Deepfake Detection Platform

#93

Post by Geee » Fri, Mar 25, 2022, 4:45 PM UTC

btbuildem wrote: post #91

Geee wrote: post #29

We have solved this problem already for text with digital signatures. There might be a way to digitally sign speech in real-time and show the signature on a display device, which is captured on video. It could be just a smartphone app or a separate device. This way every video could be proven authentic, even if they originate from unofficial sources.

How do you hash something so fuzzy as video / speech? Two different microphones picking up the same audio will result in two digital recordings that hash differently.

That's the problem. It would require some kind of new fuzzy signature algorithm, which works on information content in speech, rather than raw audio data. In this case, exact matching of data isn't important, but major changes in information, so that changing / removing / adding words wouldn't be possible. Simple way would be to just convert speech to text first with a known method, so that both microphones could work from the same source data. The display device could actually show real-time transcription of the speech along with the signatures.