Nowadays anyone with the right technology can combine and superimpose existing images and videos onto source images or videos. Deepfakes are becoming more realistic and easier to produce, offering forgery that is qualitatively different from anything we have seen before. It has been described as “fake news on steroids”. The potential to skew information and manipulate the beliefs of online communities is immense, and immensely worrying. What about building confidence with the deep fakes rising? In whom do we trust?

During the CONNECT University session on "Exploring Deepfakes phenomena", Boris Radanović – Engagement and Partnership Manager for the UK Safer Internet Centre – explained how the "generative adversarial network" (GAN) technique became an easily available software, that is being widely used to generate deepfakes. Denis Teyssou, journalist leading Agence France-Presse (AFP) Medialab and the innovation manager of InVID and WeVerify Horizon 2020 projects showcased how the InVID free verification plugin can help to source a deepfake video and identify the original images used to produce it. Can the machine learning that produces deepfakes be easily reversed to detect them, what kind of tools can be developed, or is user awareness the key?

Watch the recording and find more about the phenomena of deepfakes. 

CONNECT University
Contact information