It is hard to believe we are living in an era in which computer geeks can create fake video, pass it off as real and completely fool the masses. Such phony video and pictures, referred to as deepfakes, pose quite the significant threat to individuals, companies and even national security.
The Basics of Deepfakes
Deepfakes are videos and pictures in which artificial intelligence is creatively used to replace everything from voices to faces and physical movement. The alternative image/voice substituted for the truth is referred to as a targeted replacement. To the surprise of many, there is no need for advanced computer skills to pull off a deep fake. At this point, the deepfakes process is primarily automated. Those who understand the nuts and bolts of coding are capable of pulling off a successful deepfake that fools the masses.
Free software is available for download that serves as a training algorithm for artificial intelligence. This algorithm communicates to the artificial intelligence how faces or other body parts should be swapped as well as the part of the face/body part to be specifically altered, how to compensate for movement and even how to remedy issues with blurred images. Deepfake programs such as FaceIt, FaceSwap and GitHub run on an array of operating systems ranging from Linux to MacOS and Linux.
Identifying and Combating Deepfakes
Increasing the public’s awareness of the potential for deepfakes is essential. People far and wide should be made aware of the fact that it is now possible to substitute images and voices through the use of artificial intelligence. Though the private industry certainly has a role to play, deepfake experts agree government researchers and academics are the primary deepfake combatants. The aim is to make it too difficult for malevolent individuals to generate convincing fakes that go undetected.
A couple methods have merged to pinpoint potential deepfakes. The amplification of the color saturation in an individual’s face to reveal minor alterations is one such method. It is also possible to track video footage for unnatural blinking. Chromatic aberrations or subtle color variations around certain shapes in pictures also indicate the merging of photos.
Accessibility to Images
Part of the problem with deepfakes is most people upload their personal pictures and video to Facebook, Twitter and other social media platforms without questioning whether that media will eventually be used against them. Aside from limiting the number of pictures/videos general members of the public upload to the web, it might also help to eliminate face sets of public figures made available to the general public. Readily available data sets featuring celebrity/politician images greatly ameliorate the challenge of generating deepfakes.
In the end, the quest to thwart deepfakes will likely boil down to a race. If the media can alert the public that a certain photo or video is actually fake before it is accepted as fact, there won’t be much potential for mass deception. Unfortunately, plenty of conspiracy theorists will insist the media is lying about supposed deepfakes to protect certain political figures or celebrities. Stay tuned. We are merely at the beginning of the deepfake era. Things are about to get quite interesting as more and more phony pictures and videos are released to the public.