Re:The State of Deepfakes: Reality Under Attack. A 2018 Repo...
by barrow-rider on Wednesday January 02, 2019 @08:42AM
(#18806067)
> GAN:
Generative Adversarial
Networks, a specific kind of deep
learning algorithm that can train a
neural network to generate realistic
imagery. Whilst GANs are not integral
to creating synthetic media, they
represent the most significant
development in the how new kinds of
synthetic media are created.
In the future, our enemies will be Anonymous GANs, or GANNONs
Re:The State of Deepfakes: Reality Under Attack. A 2018 Repo...
by giorgiop on Wednesday January 02, 2019 @09:40AM
(#18806396)
That's a good question, maybe!
The way we see it: if today we are used to scan files downloaded from potentially unsafe sources, tomorrow we will get used to scan videos to check their authenticity -- check that they have not been tampered with, changing their conveyed meaning. Arguably, the same will hold for social media and video hosting websites.
While traditional malware are there to infect other piece of software, fake videos targets the human brain. This is more akin to social engineering.
Re:The State of Deepfakes: Reality Under Attack. A 2018 Repo...
by giorgiop on Wednesday January 02, 2019 @09:34AM
(#18806361)
If you know some basics deep learning, you can find source code on github to play with. You will also find tutorials made by communities of developers and users.
Otherwise, if you look deep enough on the Internet, there are already websites offering deepfake as-a-service for a rather low price.
Not all these tools were developed for malicious purposes. Quite the opposite. But today it is up to everyone of us how this tech get utilised. And while it's becoming easier and cheaper to get realistic synthetic imagery, with better tools day by day, it is not yet obvious how we will cope with fake videos in our society.
In the future, our enemies will be Anonymous GANs, or GANNONs
The way we see it: if today we are used to scan files downloaded from potentially unsafe sources, tomorrow we will get used to scan videos to check their authenticity -- check that they have not been tampered with, changing their conveyed meaning. Arguably, the same will hold for social media and video hosting websites.
While traditional malware are there to infect other piece of software, fake videos targets the human brain. This is more akin to social engineering.
Otherwise, if you look deep enough on the Internet, there are already websites offering deepfake as-a-service for a rather low price.
Not all these tools were developed for malicious purposes. Quite the opposite. But today it is up to everyone of us how this tech get utilised. And while it's becoming easier and cheaper to get realistic synthetic imagery, with better tools day by day, it is not yet obvious how we will cope with fake videos in our society.