#NewsBytesExplainer: What are Deepfake videos, how to spot them
In a world driven by social media and messaging services, misinformation can brew and spread in no time. The problem has existed for long. But now, it's getting worse as a new form of misinformation called Deepfake is spreading through online communities and marring the reputation of popular politicians, sportspersons, and actors. Here's all you need to know about it.
First popularized by an anonymous Reddit user in 2017, Deepfakes are fabricated celebrity videos created using artificial intelligence engines. They feature the face of one person in another one's video, but the thing is, this isn't anything like regular photoshopped content. They are so realistic that you won't be able to tell that the person in a clip isn't who you think he/she is.
When used with appropriate annotations/watermarks, Deepfake videos make a good tool for satire, mockery. They can show famous people doing or saying funny/weird stuff, things they wouldn't do in a real sane world. However, if the tech is misused, the same people can be shown in poor light, like how several American actresses had their faces morphed into pornographic videos.
Along with actresses, face-swapped Deepfake videos can also be leveraged to show popular sportspersons as well as business and political leaders saying or doing inappropriate stuff. To recall, just recently, Facebook co-founder Mark Zuckerberg appeared in a Deepfake video, ranting about how he controls people's lives.
At the core, Deepfake videos stem from a machine learning system called deep neural network. The system analyzes the facial movements of one person and then uses it to synthesize the target person's face making similar movements. The whole process ends with a full-fledged video showing the targeted person doing or saying things like the source subject.
Originally, creating Deepfake videos took a bit of effort, but now, the whole image-synthesis and video creation process has been expedited with basic smartphone apps that just need you to upload a photo for swapping the face in it with the one in a video.
Deepfake videos appear so realistic that people can easily be misled into believing things or activities that haven't happened at all. However, content like this could also be flagged and dodged by simply not taking any video that appears online at its face value. If a video shows something of relevance, check who has sent the clip and verify its authenticity online.
Among other stuff, you should also look for things that are out of the ordinary while trying to spot Deepfake videos. This includes any variabilities in facial features or the rate of blinking, which many algorithms fake to replicate while creating a Deepfake video. But, do note that as technology evolves, it will become more difficult to spot these elements.