Although the development of technology brings groundbreaking conveniences in many aspects of our lives, in some cases technology can also cause malicious use. In such cases, it is of great importance to be able to identify the problem, solve it and analyze the results. Here are “What is deepfake, how is it done?”, “How to detect deepfake?”, “Is it possible to distinguish deepfake contents?”, “What kind of threat does deepfake pose?” answer to your questions…
What is Deepfake, How is it Made?
Deepfake is a english word. It is formed by the combination of the words “Deep” meaning deep and “Fake” meaning fake. It is the process of adding a person to a video or photo that he has never been to, by making use of artificial intelligence. This situation can cause great problems in many respects if the person does not have permission and knowledge.
There are many methods that enable the creation of a deepfake. Among the most preferred methods is face swapping. This method, which includes deep neural networks and automatic encoders, works with artificial intelligence. It is necessary to determine a video for deepfake, and files consisting of images and videos of the person who is wanted to be included in this video are needed. The target video and the person's videos may be completely unrelated. This is not an obstacle to deepfaking. Because the automatic encoder detects images of the target person from different angles and works on matching them with the person who shows similarities in the target video.
How to Detect Deepfake?
Deepfake is a professional fraud method as it is created with artificial intelligence. However, it is possible to detect Deepfake by certain methods.
These methods are:
- You can pay attention to the eye movements of the person you think has been deepfaked. If the eyes move independently of the video environment or if the target person does not blink at all, this is an important finding that deepfakes are applied.
- Gestures and facial expressions may not match the video theme.
- Although deepfake is a method using advanced technology, the facial features of the target person cannot always be successfully videotaped. For example, facial disproportions and inequalities in skin tone may occur. Likewise, the body shape may not fully coincide with the face.
- The images of people in the videos normally differ according to the video angle and light. Deepfaked videos do not change the image of the target person in accordance with natural light and angles.
- Hair is an important factor in detecting deepfakes. In the natural flow, the hair fluctuates and changes direction in accordance with the movements. When deepfake is applied, serious changes can occur in the direction of movement of the hair.
- Apart from these observations, you can also benefit from the tools developed to understand whether the video is fake.
What kind of a threat does Deepfake pose?
Deepfake is an advanced technology as it is based on artificial intelligence. It can examine all the movements, facial expressions and facial features of the target person to the finest detail and learns the way of speaking. This makes it easy to place the created copy on the person with the appropriate characteristics. In other words, the person who becomes a target can be shown to be in an environment where he has never been before. He may be under suspicion for making inappropriate statements and may be targeted by many groups and individuals. This can lead to confusion, misunderstandings, and even legal penalties. Especially known people may be ostracized by the public or exposed to social pressure because of this technology.