Nouvelle méthode de marquage pour lutter contre les deepfakes
Paris – 29 février 2024 – Des chercheurs ont mis au point une nouvelle technique de marquage vidéo, utilisant un éclairage spécifique appelé Noise-Coded illumination (NCI), pour contrer la prolifération des deepfakes et des vidéos manipulées. Cette méthode invisible encode une image de la scène originale, permettant de détecter toute altération. La technologie, bien que prometteuse, dépend de l’adoption généralisée de cet éclairage spécial, ouvrant la voie à une nouvelle ère de vérification de l’authenticité vidéo et posant la question de savoir si elle pourra réellement stopper la vague de désinformation.
Here’s a breakdown of the article, summarizing the key points:
Main Idea: Researchers have developed a new method to “watermark” videos using a special type of lighting called Noise-Coded illumination (NCI). This watermark isn’t a visible message, but rather a coded image of the original scene, making it easier to detect manipulation.
Key Details:
how it effectively works: NCI adds a “temporal watermark” to videos recorded under its specific lighting. This watermark encodes an image of the original,unaltered scene.
Detection of Manipulation: If a video recorded with NCI is altered, the coded images within it will be unintentionally changed, revealing the tampering to forensic experts.
Potential Applications: Useful for events and interviews where preserving the integrity of video evidence is crucial. Challenge: The technology relies on widespread adoption of the specialized NCI lighting.
* Context: This research is happening as AI-generated and AI-edited videos are becoming increasingly realistic, making it harder to distinguish real from fake. The goal is to create a watermark that is impossible for malicious actors to remove.
In essence,the article discusses a proactive approach to combating deepfakes and video manipulation by embedding a hidden,detectable signature into the original recording.