It's hard to tell if DeepFake's face is fake, but Tom Cruise looks more like me than he does!

On TikTok, one of the most popular social media platforms in the world, a steady stream of deepfake videos are popping up. These videos digitally transform the subject's face or body, transforming the user into a completely different person. Of course, people usually change faces of famous people. For example, @deeptomcriuse on TikTok has released dozens of deepfake videos impersonating Tom Cruise and has 3.6 million followers.

Good guy, the similarity is so similar that even Tom Cruise himself was confused after seeing it.

In another deepfake video, a fake Zuckerberg admitted his ambition to steal user big data.

"Imagine: when a person holds the data of billions of people, he controls their lives and future... Specter asked me You know, whoever controls the data controls the future." Please remember, none of the above words were said by Xiao Zha. Look carefully at the original and reproduced videos, it is indeed difficult to tell the authenticity...

In this spoof video, Keanu Reeves has just stopped an armed robbery .

‍

And this Chinese Musk even deceived Musk himself.

It’s not the technology that’s wrong, but the people

Although most of the time, Deepfake is used by netizens to make things weird , but what is worrying is that more and more people are using this technology to carry out false propaganda, fraud, and smear public figures. Deepfake is a good blend of text, image and lens processing. It provides a cheaper and more sophisticated technology for visual insertion than green screens and computer-generated imagery. Originally, this technology was born with the best wishes of its creators, such as helping people mourn and remember their deceased relatives.

Most people would not believe such an obvious spoof. But what about that kind of specious video?

In 2019, the British Conservative Party was suspected of forging a video of Labor MP Keir Starmer slurring his words in a TV interview during the election.

And DeepNude 2.0 (now discontinued) was used to make clothed women appear naked in the camera, often used for revenge and blackmail. In addition, Deepfakes are often used for fraud. When a trusted friend or relative makes a video call asking you to borrow money, you will most likely be scammed. Fraud using digital technology cost U.S. financial institutions $20 billion in 2020, a study found.

Deepfake is easy, legislation is difficult

The creators of DeepFake emphasize that if you want these face-changing videos to look more real, you have to spend a lot of time and effort energy.

However, DeepFake is indeed becoming easier and easier. As more DeepFake applications are developed, it will become increasingly easier for people with low technical skills to create realistic-looking DeepFake videos in the future.

As long as you think about how PS has developed vigorously in the past ten years, you can probably predict the future development of DeepFake. And legislation, regulation, and detection software are struggling to keep up with DeepFake’s boom. In 2019, Facebook was criticized for failing to remove a synthetic video of politician Nancy Pelosi because it did not meet its definition of a deepfake. In 2020, Twitter and TikTok prohibit the sharing of synthetic videos that may deceive, confuse, or harm others (unless clearly marked as a DeepFake). YouTube has removed all DeepFake videos related to the 2020 U.S. federal election.

However, even if these policies are intentionally guided, it is difficult for platform administrators to process every reported video in a timely manner. In Australia, lawyers for New South Wales-based Ashurst said existing copyright and related laws may not protect Australians from DeepFake. Although foreign countries have begun to try to formulate laws to regulate this field, most of these laws focus on political communication. For example, California makes it illegal to post or distribute digitally altered candidate content during an election. But people who are not involved in politics are not protected by this law. One of the best remedies is for users to master identification skills as much as possible.

what can we do?

Often the first sign that a DeepFake is compromised is that something doesn’t look right.

If this is the case, look more closely at the subject's face and think about it: Is the face too smooth, or does it have unusual cheekbone shadows?

Do the movements of your eyelids and mouth look choppy or unnatural?

Does the hair look real?

Be aware that current deepfake technology has difficulty maintaining the original appearance of hair (especially facial hair). You can search the web for keywords about the video or the people in the video, as many questionable DeepFake videos have been debunked. Try to judge the reliability of the source – does it seem genuine? If the video is posted on a social media platform, is the poster’s account verified? You can also take a screenshot of the playing video and perform a Google reverse image search to check if the original version of the video exists and compare it with the suspicious version. Finally, if you do find a DeepFake video, remember to report it.

The above is the detailed content of It's hard to tell if DeepFake's face is fake, but Tom Cruise looks more like me than he does!. For more information, please follow other related articles on the PHP Chinese website!