One of the hot topics as the 2020 presidential election draws near is the role that fake news has played in manipulating voters. The latest scheme to hit social media in full force is deep fake videos. These videos rely on cutting-edge technology to manipulate images and audio from political figures to make it appear as if they did or said something that is completely fake. The U.S. government is concerned that deep fake videos could be co-opted as a tool by foreign governments to meddle in our next elections. Deep fakes could also be used by extreme fringe political groups.
CNN recently reported that Professor Hany Farid is working on developing technology that can root out deep fake videos before they spread like wildfire across social media. The way that the software works is that it develops a baseline profile of the facial expressions, speech patterns and body movements of top political figures that can be compared against videos posted on social media to determine whether they are authentic. Professor Farid is reluctant to share in-depth details on how his software identifies deep fake videos out of concern that those producing deep fake videos will use this information to produce even more convincing fake videos.
This technology is being tested against manipulated videos based on SNL impressions of popular politicians to establish a baseline for how accurate the software can be in identifying movements and gestures that are fake. Professor Farid says his goal is to achieve a 99-percent accuracy record in terms of identifying deep fake videos. Professor Farid’s major concern is that it does not take much manpower at all to quickly develop a fairly convincing deep fake video. He fears that they will spread rapidly at the hands of corrupt governments in an effort to generate disinformation that will confuse and manipulate the average American voter.
It remains to be seen whether Professor Farid will end up sharing his deep fake detection software with the U.S. government or will use it to start his own company dedicated to uncovering information about deep fake videos. A useful piece of information that Professor Farid’s software is expected to provide is the source of the original video that was used to produce the deep fake version. This will contribute to the base of source material that the software will study to nail down instant signs of an altered video.
Dil Bole Oberoi
