Deepfakes are nothing more than fake and altered media content. Very often these are pictures or videos. Most of the time, entire people or just their faces are exchanged. The question quickly arises as to how dangerous such deepfakes actually are.
The term deepfake is made up of the two terms “deep learning” and “fake”. The machine is always learning, artificial intelligence is used for this purpose, and networks are constantly being upgraded. Thus, it is possible that the information can be processed in several layers. This technique is also used for facial fakes, and such fakes are now even generated automatically.
This story began with the fake portraits in 2014. At that time, the faces were still invented directly by artificial intelligence. The faces did not exist in reality. On the one hand, the image was faked, on the other hand, it was checked. Image enhancements followed when this forgery was detected. Initially, it was not difficult to detect these forgeries. However, due to rapid technical progress, this is hardly possible today. Both the computing power and the image resolution ensure that forgeries can hardly be detected anymore. A new development took place in 2017. The fake images were gradually regenerated and reinvented. Both the density of the pixels and the image quality improved steadily. This resulted in images of people that looked deceptively real, even though they did not exist in real life. This was followed by moving images.
The first videos with replaced faces were published in the fall of 2017. Initially it was about pornography, but now these deepfakes are used almost everywhere. Meanwhile, even every private person can create and use such deepfakes with the help of a small and free app. There are also programs for the PC.
This allows the private user to create small video clips relatively easily and quickly. Some functions also partly allow loading and using pictures of other people from the Internet. There are also software variants that can train their own algorithm for each face.
With all these fakes, it's not surprising that some people get scared. You quickly ask yourself whether you can still trust your own eyes. The following questions may be helpful for a quick check:
· Is the situation depicted in the video actually realistic?
· Is there a concrete and reliable confirmation of the respective statement?
· Where does the information come from - is it a reputable source?
The aforementioned deepfakes have now been banned by the biggest social media operators, such as Twitter, Facebook, YouTube and TikTok. If these pictures and videos are uploaded, they are automatically deleted. However, the software for detection is still lagging behind. Soon there will be deepfakes that are not recognizable as such.
New opportunities for the movie industry
The potential of face faking has also been discovered by the movie industry. Never before has such high image quality been possible. Even deceased movie stars appear in new films; real actors only have to provide their faces for one or two days for a complete movie to be made. So in the future, it will be increasingly difficult to actually tell whether a real person is being presented on the screen or not. The potential is enormous.
Typical indications for deepfakes
There are typical deepfake hints so that you can unmask them as such:
· review of the entire context
· the face looks unnatural and too smooth due to exaggerated perfection
· the person's facial expression appears rigid and stiff
· the movements of the lips are not synchronous
· the face is partially blurred during fast movements
· the facial contours are partly blurred