Deepfake Video: The Double-edged Innovation

By Marvel Kinantan, DIGITS Staff Writer

Ever thinking of hearing wise words from your late grandma? Or want to be taught by Richard Feynman the legendary teacher? Technically it is not possible since both your grandma and Richard Feynman are gone. One possible solution is by using time machine to go to the past to meet both of them but until today it is still not possible to create one, so another possible solution is by relying on “Deepfake Technology”.

What is it?

Deepfake (a portmanteau of “deep learning” and “fake”) is a technique for human image synthesis based on artificial intelligence. It is used to combine and superimpose existing images and videos onto source images or videos using a machine learning technique known as generative adversarial network. The phrase “deepfake” was coined in 2017. So basically to hear wise advices from your grandma, you need to gather huge amount of your grandma’s photos or videos, and uses certain software to create the deepfake video and voila! You meet her once again.

How is it going?

One of the researchers in this area, Supasorn Suwajanakorn said this on his TED Talk “By the way here’s Feynman, was also known as a legendary teacher. Wouldn’t it be great if we could bring him back to give his lectures and inspire million of kids perhaps not just in English but in any language… or maybe using this tool, book authors, alive or not could read aloud all of their books for anyone interested.” He said on his talk that he is very excited due to the endless creative possibilities of this project. The latest improvement of this project per June 10 2019 is that the latest software is able to let user edit the text transcript of a video to add, delete or change the words coming out of somebody’s mouth. The improvement is achieved by Standford University’s Scientists. One of the example is they can change the famous line from Apocalypse Now woth the line “I love the smell of napalm in the morning” edited into “ I Love the smell of french toast in the morning” you can check it here.

New problem, eh?

As innovative it could be, it also brings a new threat to the digital world since deepfake videos sometimes or maybe almost all the time are pretty convincing. One of the latest case was deepfake video of Mark Zuckerberg where few artists make it looks like he brags about tricking facebook users to share their data. This kind of video can be both harmless and harmful depends on the motive of the creator. Example of the harmful one is the fake porn video of “Wonder Woman” actress Gal Gadot which was published late 2017 in Reddit by a user named “deepfakes”, which can probably harm the reputation of Gal Gadot to some extent, and many deepfake videos uses the image of High-profile figures, if it is used for evil it can harm many people or even a nation or nations. There is nothing we can do to stop the advancement of this technology and it will become more accessible to more potential users, as an example at the end of June 2019 there is a web app called DeepNude which allow users to create realistic naked images of women by just uploading photos to the app. Since the app is considered dangerous it has been shut down but it is not impossible that there will be applications similar to the DeepNudes published in the future.

Then as technology nonexpert what should we do (for now)?

Detecting Blinking

Since the algorithm for deepfake is fairly new, there is a flaw that we can see in the fake video that the person in the video blink less than a normal person. Healthy adult humans blink somewhere between every 2 and 10 seconds. This flaw happens because deepfake videos uses images of a person in the internet and it is quite rare to find images where the figure’s eyes are closed, since the photograher do not usually publish image where the figure’s eyes are closed.

Detecting Deepfake Video signs

Deep Learning specialized writer, Jonathan Hui says to look for some points in a video to determine whether the video is fake or not, first the viewer needs to slow the video down and look for these things:

Blurring evident only in the face part, while the surrounding is not
Skin tone change by the edge of the face
Double chins, double eyebrows, or double edges to the face
Face parts gets blurry when it is partially obscurred by hand or another object

Rely on detection platform

If you’re still not confident about the authenticity of the video you see using the methods above why not lend help from detection platform, one of them is called Truepic. In this platform you can upload the video to the app and Truepic will determine the authenticity of the image using over 200 verification tests such as: the intergrity of the capture device; date and time of capture; location of capture and many more. Furthermore Truepic will store the original image along its metadata, cryptographic signature, and the results of the verification tests are stored in the Truepic vault in perpetuity. Another platform that can be used is SurfSafe. SurSafe is browser plug-in that perform reverse-image search which will look for the same content that appears on trusted source sites and flag well-known doctored image, so when you click a certain image SurfSafe will do the job.

Deepfake technology is till developing and we can’t stop the development and it will become a double-edged sword for our society, as the Deepfake technology grows the methods to counter it also grows as well but in the end those methods are merely tool for us to find a “truth” which means it still depend on us to use the tools to unmask the media falsity.

Marvel Kinantan

References

https://en.wikipedia.org/wiki/Deepfake

https://www.theverge.com/2019/6/10/18659432/deepfake-ai-fakes-tech-edit-video-by-typing-new-words

https://www.businessinsider.sg/dangerous-deepfake-technology-spreading-cannot-be-stopped-2019-7/?r=US&IR=T