|Facebook launches Deepfake Detection Challenge to create automated tools that can spot fraudulent media / Credits: Alexey Boldin via Shutterstock|
While it’s not easy to create a deepfake as it requires a good amount of technical expertise, the apps and services are making it easier for anyone to learn how to create one. For instance, a Chinese deepfake called Zao offers people a simple way to superimpose their own faces onto actors like Marilyn Monroe and Leonardo DiCaprio. Cristian Canton Ferrer, Facebook’s artificial intelligence Red Team, stated that deepfakes are a growing danger not only to Facebook but also to democratic societies.
Thus, Facebook is working to eliminate deepfakes by ethically creating ones. The company launched Deepfake Detection Challenge last December, a global competition participated by researchers across the world vying to create automated tools that can spot fraudulent media. According to IEEE Spectrum, the flagship magazine and website of the IEEE, the world’s largest professional organization devoted to engineering and applied sciences, Facebook will accept entries through March 2020. It has also dedicated more than $10 million for awards and grants.
In Deepfake Detection Challenge, researchers will not only scan for signs of facial manipulation but also keep an eye on new and emerging attack methods. This includes full-body swaps that change the appearance and actions of a person from head to toe. “There are some of those out there, but they’re pretty obvious now. As they get better, we’ll add them to the data set,” Ferrer said. The deepfake detector that would be developed can be used for mundane videos as well as videos featuring popular personalities.
But, Facebook is not the only company working to detect deepfakes. The Defense Advanced Research Projects Agency (DARPA) launched its Media Forensics Program back in 2016, a year before the first deepfake videos surfaced on Reddit. Matt Turek, the program manager, stated that the researchers working under the program developed a number of detection technologies. Generally, they are looking for “digital integrity, physical integrity, or semantic integrity.”