Machine generated videos like Deepfakes – Trick or Treat?

A Reddit user named “DeepFakes” had posted real-looking explicit videos of celebrities last year. He made use of deep learning techniques to insert celebrities’ faces into the adult movies. Since then the term “Deepfakes” has been used to describe deep learning techniques that help create realistic looking fake videos or images. Video tampering is usually done using generative adversarial networks.

Why is everyone afraid of deepfakes?

Deepfakes are problematic as they make it very hard to differentiate between the fake and real videos or images. This gives people the liberty to use deepfakes for promoting harassment and illegal activities. The most common use of deepfakes is found in revenge porn, fake celebrities videos and political abuse. For instance, people create face-swap porn videos of ex-girlfriends, classmates, politicians, celebrities, and teachers. This not only counts as cyberbullying but poses major threat overall as one can create a fake video showing world leaders declaring war on a country.

Moreover, given that deepfakes seem so real, its victims often suffer through feelings of embarrassment and shame. Deepfakes also cause major reputational harm. One such example is of a 24-year-old, Noelle Martin, whose battle with deepfake pornography started six years ago. Anonymous predators stole her non-sexual images online and then doctored them into pornographic videos. Martin says she faces harassment from people till this day. Other victims of deepfakes pornography include celebrities such as Michelle Obama, Emma Watson, Natalie Portman, Ivanka Trump, Kate Middleton, and so forth.

But, Deepfakes isn’t just limited to pornography and has made its way to many other spheres. Deepfakes can also be used as a weapon of misinformation since they can be used to maliciously hoax governments, populations and cause internal conflict. From destroying careers by creating fake evidence of them doing something inappropriate to showing soldiers killing innocent civilians, deepfakes have been wreaking havoc.

In defense of deepfakes

Just as any tool can be used for good and bad, deepfakes is just an effective machine learning tool that creates realistic videos. Even though deepfakes are majorly used for inappropriate activities, some have put it to good use. For instance, GANs or generative adversarial networks (which help create deepfakes) can create realistic images of skin lesions and create examples of liver lesions, which plays a major role in medical research. Other examples include filmmakers using deepfakes for making great videos with swapped in backgrounds, snapchat face swap photo filters, and face swap e-cards (eg; jib jab app) among others.  

Are deepfakes trick or treat?

If we make pros and cons list for deepfakes, cons seem to outweigh the pros as of today. Although it has its potential good applications, it is majorly used as a tool for harassing and misinforming people.

There is a long way to go till deepfakes achieves itself a good rep and right now, it is mostly fake videos, fake images, false danger warnings, and revenge porn. Trick or treat? I spy a total TRICK!