New fake videos of actor Tom Cruise hit TikTok under the name @deeptomcruise, and man, they look real. They are so realistic, in fact, it is possible that you would not even know they are computer generated if you had not been alerted by the account identifier. And they were made using not much more than sample images of Cruise and deepfake technology that is getting easier for anyone to use.
Not even two years ago, it would have been easy to differentiate between an actual AI-generated video. But the technology is advancing so quickly that we have reached an escape velocity point, and it is obvious that deepfakery will not be used only for innocent purposes, like animating photos of your previous relatives.
This Tom Cruise video? It is a computer generated scam.
Seeing is believing – In a series of tweetsRachel Tobac, CEO of SocialProof Security, warns that deepfakes like @deeptomcruise threaten to further erode public confidence in a world where media literacy is poor and people can no longer agree on what is true or false. Like the black and gold dress, where one person may notice that Tom Cruise’s videos are synthesized, another may not know the signs of a scam and swear they are real.
“Just because you feel you can personally tell the difference between synthetic and authentic media, it doesn’t mean that we are ready to proceed,” she says. “What the general public believes is important.”
Deepfakes are especially dangerous because the video is widely considered to be indisputable evidence. A prominent individual can be tricked into committing a hate crime, or a person who has committed an unjust act can use forgery to create an alibi.
AI vs AI – As deepfakes have gotten better, companies like Microsoft have developed tools that can detect them, generating confidence scores based on signs that some of the content is being manipulated, such as the appearance of subtle fading or invisible gray elements with the naked eye. And in fact, a website called CounterSocial was able to detect Tom Cruise’s videos as fake when they were run using its own artificial intelligence algorithm.
But it is in TikTok that the videos are hosted, and this application does not have any type of integrated deepfake detector. The problem with retroactive detection is that by the time the false information is flagged, it may be too late to reverse the damage. Microsoft is trying to resolve this with a digital signature tool. The idea is for content creators to sign a video’s metadata as it is being encoded, and a tool that users install in their browsers would alert them if the metadata has been altered in any way. It is still an experimental concept.
Label time – Tobac says applications like TikTok need to have integrated deepfake detection and labeling tools. And until then, Cruise and others will need to check themselves out on these social networks just to get ahead of counterfeits.
Even after these detection layers are added to applications like TikTok, deepfake technology will continue to improve, and that probably means a constant game of cat and mouse going forward. Microsoft admitted that even its detection technology will need to be constantly updated. That is simply the nature of technology.