AI vs. Blockchain — Solving the Deep Fake problem using blockchain, IPFS (and NFTs?)
Can Blockchain help prevent the spread of misinformation caused by the exponentially growing AI's ability to mimic the real world in both text and images?
I’ve been thinking about the deep fake problem for over a year. But, I must admit ChatGPT caught me off-guard. I knew AI was coming (not just ChatGPT, but many others we do not know about yet), but I did not realize how fast it would be here. And now that it is here, I am worried things might get ugly quickly.
Not just because of deep fakes, but a perfect storm might be brewing for AI, given its capabilities to replace humans in specific job roles combined with the global economic outlook and its effect on the job markets worldwide.
So, last night I was backed into a corner by my internal alarms, and I’ve decided to put some of my thoughts into writing and take action on them.
This article is about an imaginary blockchain-based product/technology/approach named Testimonio that could help protect us from the images and videos generated by AI that are becoming increasingly difficult to tell if they are fake or not.
The Problem
The simplest definition of a problem we face is how to prove a photograph is real and represents real events that happened in the real world.
The Solution
Introduction
So, what constitutes a solution to this problem, and what exactly do we need to know to say or assume a photograph is real?
If we knew when the photo was taken, where the photo was taken, by whom, and by which device the photo was taken, we could say that the photograph originates in the real world.
When I say “by which device,” I do not mean is it iPhone 6 or iPhone 14 — I mean the serial number of the device that can consequently be linked to a person who owns the device and linked to other hardware the device is made of, where the device was made, by whom and when.
For those of you who already see where I am going with this, the solution I will be describing in this article is multi-phased. It starts with mobile devices/cameras but expands into professional cameras (Canon, Nikon, etc.) and image processing tools such as Photoshop.
The concept
Each image-producer device has a unique private key that allows it to sign transactions and interact with the blockchain.
Upon capturing the image or video, the image-producer device will collect the metadata about the media file, including location, time, device’s hardware identification numbers and other information about the device, and checksum of the media file.
The original media file is then stored on IPFS. Finally, together with the rest of the information previously collected about the media file, everything is recorded on-chain by signing and sending a transaction to the Testimonio smart contract.
Each media file the Testimonio smart contract records on the blockchain is given a unique identifier. When queried with this identifier, the smart contract will provide you with the original media file (its IPFS location) and recorded information/metadata.
The implementation
Mobile devices — proof of concept
The Testimonio proof of concept will come as a mobile application that, like any other app, can be downloaded and installed via an app store.
Upon installation, the app will generate its’ private key for interaction with the blockchain and register the device with the Testimonio smart contract. In exchange, the smart contract will airdrop a certain amount of tokens to the user, to pay for future transaction processing costs.
Those familiar with the blockchain are familiar with the concept of “gas” and know that processing transactions on the blockchain is not free, but I will return to this topic later in the article.
The mobile application allows users to take photos and videos. When a media file is captured, the app will collect and prepare the metadata, and the user can then record it on-chain directly from the app.
Only photos and videos taken from the Testimonio mobile app can be recorded on-chain via the Testimonio smart contract.
If the blockchain is our truth-keeper, we need to gate and protect access to it. Hence only the photos taken directly from the app can be considered truthful in this case.
Image processing software — Phase 2
Support for image processing software is crucial, and this would come in the second stage of the Testimonio development.
Coming in the form of a “Photoshop plugin,” the end user could record post-processed photographs or videos on-chain, given that a reference to the original file is included.
The “Photoshop plugin” would interact with the blockchain and the Testimonio smart contract the same way the mobile application does — the plugin would be registered with the smart contract, and the plugin would have its private keys for signing blockchain transactions.
The idea is not to prevent image processing but rather to have transparency, awareness, and verifiable origin of the processed image, and thus the proof that the processed image still represents the actual events that happened in the real world.
Professional cameras — Phase 3
Although most photos and videos taken today come from a mobile device, it is essential to have professional cameras enabled with this technology, especially for the press and professional photographers.
Existing or new blockchains
The proof of concept will be based on an existing public blockchain such as Ethereum or Polygon. But if this idea is proven to have merit, the question of using a private blockchain or having its own blockchain would have to be deeply considered.
The goal would be to find a balance between decentralization, cost to the end user, and operating costs.
What’s in it for The Press and the photographers? — Phase 4
I can’t escape the thought that technology, such as what I described above, opens the doors for the “Istokcphoto X.0”.
The blockchain would allow licensing photographs and videos and providing proof of the license to be much easier than today.
Imagine reading an article on CNN, this technology would allow you to know that the photos you see or the videos you watch in the article are not fake or AI-generated, and it could be easily verifiable that CNN has the right to use the image.
Although this last part would be of much more interest to CNN or the professional photographer who took the photo because they would want to know if XYZ portal has the license to use their images.
Conclusion
Thank you for reading this article. I did not intend to write a full-blown technical specification but rather to spark your imagination and inspire you to think about the problem.
Although I can not tell you when the Testimonio proof of concept app will be available, please sign up for the mailing list here if you want to be notified.
I would like to hear your thoughts on the problem and the solution I described, so feel free to comment or DM me.
New to trading? Try crypto trading bots or copy trading on best crypto exchanges
Join Coinmonks Telegram Channel and Youtube Channel get daily Crypto News
