The BBC’s mini-series The Capture makes a strong, if indirect, argument for blockchain technology.
The show depicts a world full of fake news, where intelligence agencies use technologies like deepfake face-swapping to manufacture evidence supporting their agenda. This is not science fiction; thanks to the GAN (Generative Adversarial Network) that emerged in 2014 as well as open source technology, deepfake videos are one of a growing number of ways people can create convincingly realistic “fake news.”
Deepfakes are already being used for everything from jokes to porn. But as I wrote in a previous article, that the real danger in deepfakes is the potential to use them for fraud and terrorism. This is precisely what the show depicts. In The Capture, even “reliable” CCTV surveillance video can be altered, asking the question: can we no longer believe what we see?
As far as anyone knows, we haven’t seen technology as sophisticated as what’s used in The Capture, which depicts realistic deepfakes being generated in real time, show up in real life. But it is not difficult to imagine that such technology could exist. Real-time environment capture, deepfake neural networks and motion libraries all exist. Perhaps no one has combined them to create totally convincing real-time deepfakes yet, but it will be possible someday soon.
And while it seems incredibly advanced and out-of-reach, it could quickly become normal and ubiquitous. Under the auxiliary of various tools that are easy to use, people can quickly learn the technology and put it into practical use.Hao Li, an associate professor of computer science at the University of Southern California and an expert in facial recognition whose work on performance-driven facial animation contributed to the iPhone X’s Animoji,said in an interview with CNBC this September that everyone will be able to create manipulated images and videos that seem “perfectly real” in “half a year to a year.”
Indeed, they already can. Deepfake apps already put this technology into the hands of the public, although the results aren’t often flawless enough to be indistinguishable from the real thing.
This is a big deal. As Tom Van de Weghe of Stanford University pointed out in his blog, “[Deepfakes] can be used to create digital wildfires. They can be used by any autocratic regime to discredit dissidents.”
As long as data exists, data tampering exists. So in an increasingly data-driven, digitalized world, what can we believe? This is where blockchain, which has the potential to serve as a powerful anti-tampering tool, comes in.
Data recorded on a blockchain is incredibly difficult to tamper with, since information is verified and stored by a large number of decentralized nodes when it’s entered into the chain. Changing that information after the fact, therefore, would require controlling at least half of all available nodes — a hugely difficult and expensive prospect on most blockchain networks.
And while storing entire videos on the blockchain would be resource-intensive, that wouldn’t be required. Only the fingerprint-like hash value of the video data for each second (or perhaps each frame) would need to be recorded. Since these hashes are relatively short strings of characters, they’re much smaller to store than raw video, but they can still be used to identify any video that’s been modified from its original.
For example, here’s how this approach could be used to defend against attacks on security camera footage: we can add a signature to each camera with a unique private key built into the chip.The camera then signs each frame/second/minute of the footage using this key, and the video data hash and camera signature are recorded on the blockchain. And any data tampering could be quickly discovered by generating a hash of the potentially-tampered file and comparing that to the original stored on the chain.
A similar approach could be taken for the video recorded on mobile phones. Many modern phones already have a chip or dedicated chip area reserved for security. For instance, Google’s new product Pixel 3 is equipped with such a security chip named “Titan M”, iPhone has similar ‘Secure Enclave’, and Samsung's Galaxy and other Android phones often use TrustZone that comes on ARM chips.
Right now, these security chips are storing data internally, but they could be used to generate and store hashes for videos on a blockchain, similar to the approach described above for security cameras.
As the world becomes more and more digital, the need for counterfeit-proof digital technology is likely to rise, and blockchain is the best answer we currently have. It can offer security and certainty, freeing us from our reliance on corruptible centralized institutions to determine what’s true.
Authorities as varied as Facebook, Google, and the British government are trying to fight deepfakes in their own ways, but any such centralized solution requires trust in the institution itself. Do you trust Facebook? Personally, I prefer the distributed, trustless solution offered by an immutable blockchain — a single, transparent, and distributed source of truth that anyone can verify for themselves.