What Is Deepfake And How Can It Be Detected? | Use Case | Detection |

Hello and welcome back to yet another blog, and today we are going to discuss one of the most exciting technology that is coming up, and it is none other than Deepfake

The internet has always been a bit of a minefield when it comes to finding correct information, from reliable sources and reputable people. So, as we look into the deep neural networks, machine learning, and devious minds, that prepare Deepfakes. This technology takes one person's face and replaces it with another, in an image or video. This is the Future of Tech. While the word Deepfake sounds mysterious and straight out of a spy novel. What does it mean?

What Is Deepfake

The word Deepfake is made up by combining Deep Learning and Fake. With Deepfake, you can switch any footage of a person's face with another person's face. Deepfake isn't only about manipulating video though. There are also examples of audio fakes, where you can create any sentence said, in a specific person's voice. For example, you could make your colleague say that you are the coolest guy in the world! isn’t that cool.

Use Cases Of Deepfake

While Deepfakes is kind of fun and impressive. What are some of the real use cases for them? Let's start with the most obvious that you may have seen, I.e... manipulating video footage. Often this has seen with something silly, like a celebrity face swap, or putting one person's voice on someone else's face, but it goes much further than that.

You can also bring back actors from the dead. If a sequel to a popular sci-fi franchise, for example, needed to use those actors, again. Museums are looking at bringing back to life, historic characters, such as David Crockett, John F. Kennedy, or even Gandhi. Having historic events recreated with the real people and voices would enhance the learning experience dramatically.

A clinic in Boston, in the US in conjunction with Northeastern University, is recreating voices for people that have lost their own. This pioneering project can give people an important part of their identity back, which is cool.

Of course, it isn't only the good guys, that get hold of amazing technology. In one instance, a UK energy company's Chief Executive has tricked into wiring 200,000 € to a Hungarian supplier, because he believed his boss was instructing him to do so.

Political Effects

Imagine you could make any politician say, anything you want? A famous example is Barack Obama saying things he would never say, well, at least not in public. It was done by a comedian Jordan Peele, who mimicked his voice, and Deepfake technology then applied Obama's face and the nuisance was created.

A more subtle example is Indian politician Manoj Tiwari, who used Deepfake to make himself appear to speak fluently in the dialect of his target voters. Now, does that mean we can trust what politicians are saying anymore? Wait, never mind, we never did so!

Ethical Implications

The real threat is the use of this technology to spread misinformation. The ethical implications and considerations for technology, like Deepfake, are vast. Imagine this: A criminal is charged with a crime, the criminal's defense, they find a video from some surveillance footage, from a place in a completely different city. The footage is then put through a Deepfake process, to put the criminal's face and person, into the footage, and that is it, all done.

Or what if phone calls could be manufactured with the exact voiceprint of a different person, that could lead to Identity Theft, Fraud, and so much more. So how does the technology work? And can we in fact harness it.

Tech Behind Deepfake

The technology behind Deepfakes is a subdiscipline of machine learning, known as Deep Learning. Deep Learning uses something called neural networks; a design that was inspired by how our brain cells work. To create a Deepfake we need to teach our neural network, how to create something completely new. We can do this through a process called training, which involves feeding in lots and lots of examples, of the thing we want to ultimately recreate.

Now after much time and number crunching, our training process will provide us with a set of instructions for how to create that new thing. It's this thing that we call our model. Now, this model is ultimately used to generate entirely new videos, pictures, or audio, based on the examples that we provided.

Training Time

So how long does the training process take? Well, models can take anywhere from 12 hours to 12 weeks or more, depending on the quality and quantity of examples, the complexity of the subjects, the computing power available, and how much money you're willing to throw at the problem.

Now, if you want to know more about Machine Learning and Deep Neural Networks, stay updated with the page as it will be uploaded soon.

Detection of Deepfake

Now, as we've seen, Deepfakes hold the potential to cast doubt on the legitimacy of any form of digital media. So how in the world are we to believe anything anymore?

Last year, a consortium of over 100 tech companies, including AWS, Facebook, and Microsoft, created something called a Deepfake Detection Challenge, offering 1 Million dollars in prizes, for anybody able to create an effective way to detect Deepfakes.

Now, another idea that's being kicked around, is some form of a digital watermark, issued and stored using a blockchain. When the digital asset is created, it's assigned a unique ID. That's a sort of fingerprint generated from the original. This could then be used to verify the authenticity of that asset later, or uncover fakes.

What's Blockchain?

Simply put, Blockchain can be described as a system of recording information in such a manner, that it becomes difficult or impossible to change the data, ensuring security, transparency and decentralization.

I expect that there will be two main streams of future Deepfake tech. One side will be enhancing the use and believability of it. The other will be detected, what is a Deepfake, and what isn't, both will be equally important. 

Conclusion

Deepfake is a great evolving technology and can be very helpful in the near future, however, it can also be used for some notorious activities. Well, it completely depends on how we look forward to it, and what measures do we take to avoid the consequences. Having an in-depth knowledge of Deepfake can land you to many Tech Giants involved in the research of it. Hope you enjoyed reading the article and as I always say, you should always have your hands-on on new tech if you are interested. Make google your Best Friend. On this note, I would like to wrap-up this article, Thanks for Reading.