Proof-of-existence for one-off documents is super awesome, and it enables anchored security models like Keybase‘s and, in a much less simplistic way, Blockstack‘s. And many others. Sometimes, you can get a ton of utility from putting only one small thing in the blockchain.
I’ve been thinking recently that, instead of putting just the small anchors of security into the blockchain, we should be putting damn near everything into it, especially media. Follow me down this rabbit hole…
There are four scenarios of real/fake digital document validation:
- real documents from the past that you’d like to prove today are real. This might be a birth certificate or a photo of yourself with Shaq from that time you actually met him.
- real documents from the past that you (or someone else) would like to claim are fake. This might be photos that an investigator took of you shtooping someone, or an audio clip of you on Access Hollywood saying that women let you grab them by their genitals.
- fake documents from the past that you’d like to claim are fake. This might be a deepfakes porn film in which someone has plopped your face onto the body of the star.
- fake documents from the past that you’d like to claim are real or historical. This might be doctored security camera footage that places you away from the scene of a crime, or a contract written after-the-fact, or a photoshopped pic of you with Shaq that makes your friends super jealous.
Our basis for understanding these four scenarios is changing extremely quickly — right now, no matter what we do, this is changing. As more of our documents and media become digital, and as the technology for manipulating digital realities improves, the cost to create convincing fakes is approaching zero. Without a significant cultural shift, all of these scenarios will start to bleed together because, for any historical document of significance (or a fake thereof), there is someone who could myopically benefit from claiming its falseness or its truth. This is especially unnerving in light of the recent DNS attack. Imagine going to CNN.com and seeing a totally convincing video of our president declaring war on China. You would have almost no way to know whether or not this was real.
We have been losing our basic ability to make deductions on what we see and hear in the world around us. I wrote about this previously: Strong Deductions and Bitcoin. This is happening. The sooner we accept that a video of the president might or might not be true, the sooner we can move towards building infrastructure that improves the situation.
We can timestamp everything. Prove the existence of everything.
If our expectations can be shifted from “what the hell is a proof-of-existence” to “what a delightful surprise that this document/video/photo has a proof-of-existence” and then eventually to “why doesn’t this document have a proof-of-existence,” we would be much better prepared to deal with the coming scenario-merge. There are subtle wins from growing skepticism across the board, but the most obvious benefit is in scenario 1. That real documents or historical media from the past can be proven to have existed before a certain date cannot be argued to be a fake. The historical immutability of the Bitcoin blockchain is just about the greatest guarantee that can be made in human-computer interactions. All we would need to do is prove the existence today (before it’s very easy and very cheap to create convincing fakes) of everything. And the surpring thing is that it might not actually be so expensive to do it.
Proof-of-existence can be really cheap (unless you use the hilariously overpriced poex.io for some reason). With merkle trees and a little offline storage, you can timestamp millions of documents into a single Bitcoin transaction (which will set you back $0.45 as of 1 May 2018). I’ve written some code (https://github.com/xgess/timestamp_all_photos) to hash and timestamp all of the photos in my Apple Photos library. It was about 15000 items, and the merkle root is now on the blockchain. An extremely similar method could work for the New York Times to hash and timestamp all of their historical photos, videos, articles, and audio clips. And a similar method could be applied to all newly generated content moving forward.
After writing my little library, I discovered a service called Tierion that offers an API with an incredibly generous free tier for most of the functionality. It’s never been easier to prove the existence of everything you have. And there’s some urgency here. We don’t need to solve all of the end-user experience problems any time soon, but there is an obvious benefit to getting this piece of the puzzle started soon. Metadata will matter more than data, because it will be easier to trust.
There is a big problem coming, and proof-of-existence isn’t going to solve it, but I suspect it could help a little. And no one seems to be proposing anything else that makes any sense.