DeepFake – A Destructive Technology

We are entering the digital era of “AI-powered by Deep Learning”. This disruptive technology promises innovative solutions, fresh markets, and unexpected ways to connect with customers and the market. The machinery is growing so the risk to society. DeepFake technology is emerging as a threat to the functioning of government, fundamentals of commerce, and social structure.

The movie faceoff is slowly becoming a reality. Blending the human image on one-another is known as DeepFake. It is swapping out on the image of a one-person face with another person using disruptive technology, “Artificial Intelligence”.

To demonstrate the potential destruction, the DeepFake technology can do, Jorden Peele creates this fake video of Ex-President Obama to show how easy it was to put words in someone else’s mouth. (More Read: Synthesizing Obama: Learning Lip Sync from Audio)

What is DeepFake?
As per wiki ”DeepFake is a synthetic media in which a person in an existing image or video is replaced with someone else”. 

As described earlier in this post, it is swapping out on the image of a one-person face with another person using disruptive technology, “Artificial Intelligence” powered by deep learning and autoencoders or Generative Adversarial Network (GAN) framework.


How is it done?

GAN Deep Learning framework is the major technique to create DeepFake videos. GANs use two types of Neural Network. The “Generator” and the “Discriminator”. The generator creates an image, and the Discriminator tries to detect whether the image created by the generator is fake or not. The generator creates a realistic image until the Discriminator fails to detect it is fake.

How scary it is?
Radicals, Criminals, and Rogue Politicians can use this to cook a fraud social campaign and damage someone’s reputation easily. Fakes are easy to create, authenticity also becomes easier to deny. It is an opportunity for the politicians to deny whatever they said earlier and claim it is not them. Meaning, politicians now can say anything and if you get them doing genuinely objectionable things, could claim evidence against them is bogus. 
Edgar quotes “DeepFake content is building chaos among people who grew up with the idea that seeing is believing”. Photos and videos, once considered undeniable evidence of something happening, is now being questioned by the masses. 

Earlier, DeepFake technology widely used only for pornography and shared in Dark Web (Deep Dark Internet). Now it is posing a great threat to society.

Let’s take a few scenarios to better understand the destructiveness of DeepFake Tech:

• Creating a video of a politician engaging in bribery or sexual harassment before the election
• Creating a video of a religious leader expressing prejudice statement over other religions

Even though it could be proved as a fake video, the damage is done already. It can make effects in society by:

• distorting democracy
• creating social/community distress
• manipulating elections
• making to lose trust with anyone such as Government, company, individuals, religious or community groups, celebrities, family members, and political parties

Rapid development in DeepFake technology is realistic, difficult to detect, and easy to make. It is very hard to detect DeepFake videos. It is hard but detectable. How?

With the current technology, the make the DeepFake video is by doing frame-by-frame independently. This is a weakness of the DeepFake tech. So, there is a smoothness problem (flicking). So, the DeepFake video makers blur the image across frames and the Technologists can design software using AI Technology powered by Recurrent Neural Networks (RNN).

Conclusion
For now, the only way we get the truth is by being more vigilant with what we trust from the internet. It is getting harder and harder to trust our eyes and ears. It is hard to determine the veracity and distinguish between legitimate and fake. For now, looking at the source, double-checking with google may solve the problem to a certain limit.

Cheers,