Deepfake: fun or fraud?

What are deepfakes?

Deepfakes, named after the deep-learning AI algorithms that function as the underlying technology behind it, are a new kind of video and audio phenomenon featuring realistic face-swaps and audio-swaps. In short, a computer program finds similarities between two faces and puts one over the other. The more of the source footage the algorithm has, the transformation is better and more seamless.

These videos were first experimental and made to amuse people but the question now is, with a lot of fake news going around on the internet are deepfakes fun or fraud? And even though it was mostly used on celebrities of every kind, it didn't stop there. Deepfakes were also used to "fake" videos and speeches of politicians bringing confusion and potentially harm.

Deepfake technology is being made easy to use. If someone wanted to make a deepfake video they could easily use the Amsterdam-based startup which published an audit of all the online resources that exist to help people make deepfakes. People can now even create deepfakes with just typing whatever they want their subject wants to say with the help of Lyrebird.

If you're asking whether there's an app for that, the answer is there is an app for everything as well as making deepfake videos. The Chinese app Zao enables users to insert their or anyone's face into for example popular TV shows or movies with just a single photograph.

How exactly the deepfake technology works?

Deepfake technology was invented in 2014 by Ian Goodfellow, a Ph.D. student turned Apple employee. Most of the technology is based on what is called generative adversarial networks (GANs). GANs enable algorithms to move beyond classifying data into generating or creating new content like images or audio. The GAN algorithm involves two separate AIs, one that generates for example photos and an adversary that tries to guess whether the images are real or fake, according to Vox.

This can be explained more simply as two GANs or AIs competing between themselves to fool each other into thinking an image is "real." Over time, each type of AI gets progressively better by learning more, and produces content that looks perfectly life-like.

Below is the more detailed explanation of the deepfake technology:

How deepfakes affect payment and fraud?

With the deepfake creation technology becoming more easy to use and more accessible, it was just a matter of time when it will be used for payment fraud.

According to the Biometric Update, deepfake claimed it first victims in August this year "with a British energy defrauded of nearly a quarter-million dollars through a wire transfer ordered by what seemed from the voice to be a company executive".

What happened was that the CEO of the company thought he was speaking on the phone with his boss, the chief executive of the firm's German parent company, who asked him to urgently send the funds to a Hungarian supplier, according to the company's insurance firm, Euler Hermes Group SA. The same insurance firm said the fraud was done by using a software that can mimic voice, tonality and punctuation. In the end, no suspects were unfortunately identified, according to the Wall Street Journal, which could pose a big law-enforcement problem in the future as AI crimes have not yet been deeply regulated.

If this happens with voice mimicking, will it affect voice payment market? Luckily, governments and fintechs are already working on developing programs to prevent this kind of fraud. Amazon, Twitter and Facebook have taken steps in deepfake detection. Amazon Web Services has announced it is joining Facebook, Microsoft, academics and other experts to encourage innovation in deepfake detection.

In this joint effort, Facebook is investing $10 million to develop technology aimed at detecting deepfakes, according to The Hill. Both Faceebok and Amazon are inviting researchers to use their datasets to build their own deepfake software to submit for the challenge.

Tags: AMAZON • DEEPFAKE • FACEBOOK • FRAUD • SECURITY • SOCIAL MEDIA