Keywords:  AI and Machine Learning, Social Engineering/Disinformation,

2017

Identifying Audio-Video Manipulation by Detecting Temporal Anomalies

Alexei Efros, Associate Professor, Department of Electrical Engineering and Computer Science
Andrew Owens, Postdoctoral Researcher, Department of Electrical Engineering and Computer Sciences, UC Berkeley

Inexpensive recording devices, from cellphones to surveillance cameras, have made audiovisual data ubiquitous, while commercially-available software has made it significantly easier for people to manipulate/alter this data. It has become increasingly important to develop tools for verifying the authenticity of audiovisual data. This research will explore a new method, based on deep neural networks, for detecting fake or manipulated videos. The method will work by identifying situations in which the audio and visual streams of a video are misaligned, a common result of video manipulation.