Mr Peter Edwards

Research project: A Deep Learning Methodology for the Automatic Detection of Deepfake Media

Abstract

Society has reached a tipping point. The loss of trust in media outlets, the spread of misinformation campaigns by governments and criminals, and the rise of misrepresented content on social media have fuelled a sense of unease and confusion. How can society judge if something is real or fake? The rise in deepfake technology has elevated the seriousness of media manipulation to an unprecedented level, with the potential for dire consequences.

To address this issue, this research aims to develop a Deep Learning (DL) methodology for the automatic detection of deepfake media. At present, many of the state-of-the-art techniques proposed in academia achieve high accuracy when tested with media in controlled environments; however, this success does not translate to media captured in the wild, therefore failing to achieve their intended purpose. The primary goal is to enhance the existing techniques by finding a more generalised solution that can be more effective at detecting deepfake media in real-world scenarios.