A deepfake is a type of manipulation that uses artificial intelligence to create fake videos or audios.often very realistic. The aim is to manipulate public opinion, disinform, damage someone's reputation or perpetrate fraud.
👉 Types of deepfakes
There are two main types of deepfakes:
- Deepfake video : creation of fake videos in which a person is shown saying or doing something they have never done.
- Déepfake audio : creation of false audio recordings in which a person is imitated, giving the impression that he or she has spoken certain words or phrases.
How it works
Deepfakes are created usingdeep learning :
- Data collection The attacker recovers a large number of images and videos of the targeted person.
- AI training A neural network is trained on the basis of this data to imitate the person's expressions, voice and movements.
- Creation of deepfake The AI generates a new video or audio in which the person is manipulated.
💥 Consequences of a deepfake
- Disinformation Deepfakes can be used to spread false information and manipulate public opinion.
- Damage to reputation A person can be discredited by the broadcasting of false videos or audios depicting them in compromising situations.
- Identity theft and fraud Deepfakes can be used to impersonate people and carry out scams.
- Cyberbullying Deepfakes: humiliating or defamatory deepfakes can be created with the aim of harming a person.
💉 Protection and remedies
It is difficult to detect a deepfake with the naked eye, but certain techniques can help:
- Video analysis looking for inconsistencies in movement, expression or lighting.
- Checking sources Ensure the reliability of the video or audio source.
- Use of detection tools AI-based tools are being developed to detect deepfakes.
- Raising awareness To inform the public about the risks associated with deepfakes and encourage critical thinking.
📊 Figures for France and worldwide
- According to a 2023 study, 96 % of deepfakes are used for non-consensual purposes, including the creation of fake pornographic videos.
- The number of deepfakes in circulation doubles every year
- The number of online deepfakes doubled every 6 months between 2019 and 2022.
