EPFL OpenDays: The Dangers of DeepFakes

Since their emergence DeepFakes, or more specifically Deep Learning-based fake media with manipulated facial identity and expression, have brought about serious security and privacy concerns. Whilst bringing their own advantages to photo and video editing, they come with enormous downsides, such as identity abuse and the spread of misinformation including defamation and political misuse to video-call scams and pornography. This has become an increasing concern because in modern society facial recognition is crucial for identifying people, either as a thought process in the brain or as an algorithm in machines. We work as a small team within IVRL at EPFL to develop DeepFake detection methods, allowing for fake media to be recognized early and to be dealt with accordingly. At this demo you will have the opportunity to see and understand how DeepFakes are made, how they present themselves, and recognize signs that can help you determine a video's authenticity.