Deepfakes and Ethics
What Are Deepfakes?
Nowadays, Deepfake videos, audios, and images are very popular. Deepfakes are fake videos, images, or audios which look real.
According to the definition, Deepfakes are digital media where a person in an image or a video is replaced with someone else’s likeness by using machine learning and deep learning techniques. It has three steps extracting, training and creation. Deep learning techniques require a lot of data so, deepfake uses video clips with the feature the people who want to face swap.
In the extraction step, extracting all the frames from video clips, identify and align the faces. The training step trains each aligned face using a neural network to convert the face to another face. In the last step, creation, merge the converted face back with the original frame. The final step does not use machine learning techniques and the algorithm is hardcoded where it lacks flexibility.
The key difference between deepfake and older face-swapping techniques is face-swapping only works with preexisting pictures. If we morph A with B, it cannot match the expression of B. on the other hand, deepfakes can morph a person’s face to imitate another person’s features while preserving the original feature expressions such as talking, smiling, and emotional expressions.
Deepfakes for good
deepfake technology is now widely available and makes realistic videos and photos. it is now hard to distinguish between real and fake media. Deepfake is a valuable technology when used with the proper training and good intention.
This technology is very useful in the movie industry especially to create characters of the dead artist. For example 2016’s Rogue One: A Star Wars Story.
Also, researchers used deepfake technology to create historical personality animations. Below is a video of the Mona Lisa, where she moves her head, eyes, and mouth.
Another huge advantage of deepfake is financial savings. This technology allows the cheap creation of movies, videos, advertisements, and games. So, this is very beneficial to the companies and individuals to enter the industries with fewer investments.
This technology also allowed Identity protection, where AI-generated avatars use to protect the interviewees in news reports. These AI avatars also use in training videos.
Threats of Deepfakes
It is common that people use technology in evil ways and there are negative effects of Deepfakes as well.
Some people use deepfake technology to take revenge by creating phonography or sensitive content. This technology can make videos where anyone can appear to be doing or saying anything anywhere. With current technologies, most of the videos are appear as real videos. Find the difference between real and fake is very hard. Especially, public figures such as Barack Obama, Mark Zuckerberg, and Google’s CEO have been targeted recently with deepfake videos.
People use this technology to pretend someone else by cloning the voices and do frauds and scams. These are critical issues in the use of this technology. Since anyone can access these technologies simply using mobile apps these frauds are very common nowadays.
How to detect deepfake media
There are some signs to detect fake from real in deepfake media.
Deepfake media can identify by paying attention to the face. There are always fake manipulations in facial transformations in deepfake media.
Another way is to check the cheeks and forehead. Deepfake often incompatible on some dimensions such as skin appear too smooth, or the age of skin is different from age in eyes or hair. Also, facial hair can use to identify real from fake. Deepfake media often fail to make facial hair transformation natural.
Deepfake normally fails to represent the natural scene perfectly. By checking shadows and eyes we can identify the real one from fake media.
Normally deepfake media shows glasses with too much glare. And by checking the angle of glare change according to the movement of the person we can identify the fake media.
Also, by paying attention to the facial moles, eye blink, and size and color of lips we can identify the fake media. Another way to identify fake media is the voice.
But with time, deepfake technology will evolve, and these issues will be fixed. Then it will be hard to identify fake media from these features or flaws. It will become impossible to detect fake media from the real one and it will cause big ethical issues.
With the growing technology, in the future, it will become impossible even for experts to distinguish synthetic media from fake media. There will be huge ethical concerns with the deepfake media if it became impossible for experts to distinguish synthetic media from authentic media.
With the existing mobile applications, it is easy for anyone to create deepfake media. We cannot restrict the use of deepfake technology, but we can add some regulations to the media.
Ethical issues are rise when creating face-swap videos without the consent of people in the video. As an example, a video clip while taking at a party with friends can use to make a deepfake video for fun. But if that video share on social media without the consent of the people involved or use that video to make phonography deepfake video or embarrassing videos for revenge is a critical ethical issue. These fake videos can go viral, and people involved with these videos will face social and emotional issues.
Deepfake attacks are mostly targeted at public figures such as Politicians, Celebrities, Business leaders, etc. There is no difference between using images of public figures and private individuals. But it is easy to find data from public figures because a lot of their videos and images can be found easily on the internet. Creating deepfake videos such as pornography, embarrassing videos, or criminal videos of public figures or individuals’ will affect the victim’s career, name, social life, and mental health.
Common examples of harmful deepfake attacks on public figures are,
- Barak Obama insulting Donald Trump video
- Mark Zuckerberg’s harmful speech about facebook’s power.
- Democratic House Speaker Nancy Pelosi’s drunk/ill while in the TV program video
- Kim Kardashian rap video where the lyrics of the rap are harmful to the person.
These things lead to suicidal and mental breakdowns for a lot of people regardless of they are a public figure or a private individual. Deepfake is a very beneficial technology if someone has the consent of people who are involved in the video and uses deepfake for a good purpose.