Many social media users are quite familiar with the iron man trend where a person replaces Robert Downey Jr.’s face with his/her. Other similar trends include Joe Biden or Vladimir Putin giving speeches related to product advertisements. Deep Tom Cruise has taken over TikTok.
People haven’t much research about it but most of us are somehow aware of how deep fake is taking over the internet slowly. To describe in brief, deepfake is an artificial neural network method to replace one person in an image/video with someone else and display the real output. It has enormous usage in entertainment sectors like karaoke impressions of people’s favorite singers, audio dubbing in many languages, and editing facial expressions. However, it has brought major challenges like cybercrime, privacy invasion, and social invasion.
Video: Deepfake Tom Cruise
DEEPFAKE & CHALLENGES
1. User Authentication
Modern websites use several new methods to provide the best authentication to users and stop the theft of personal data. They first came up with a simple username, password, email, and phone number-based authentication. Then the chat and software bots took over to access the user profile. The easy solution was using I am not a robot captcha-based authentication where the requests of a bot could be differentiated from the request of a human-based on several requests per time humans and the bot could provide to the server. These still had limitations like humans forgetting the passwords. So, fingerprint-based authentication and facial recognition were brought in. However, these two best methods soon to be applied in the future were clearly demotivated with the introduction of deepfake as all these authentications are possible with an accurate deepfake algorithm that could imitate both fingerprints and faces.
2. Social Prestige
In the initial stage, deepfake had already allowed users to create many fake, irrelevant, and immoral videos that are attacking people’s respect and personal information. Although many websites especially Reddit wouldn’t verify the deepfake files but other newly introduced platforms are slowly becoming centers of sexually misinterpreted videos with big viewership destroying the prestige of the innocent public. This is the most dangerous method which would join the long queue for sexual crimes in underprivileged nations.
Ten years ago, in a world without the internet and pirated movies, entertainers could only be viewed upon television or film halls. Nowadays, the pirated movies are taking film profits, the reason why the list of highest-grossing movies based on inflation includes very few films from the present decades. The local television channels could save loads of money with advertisements without paying the famous performers. This illegal practice could easily be introduced as the next-best crime in the entertainment sector after pirated cinema and TV content.
In journalism, the reporters could create fake videos of politicians where leaders could be heard giving falsified quotes about other people and parties. In the country’s next election this could easily affect the event and ultimately the voting results if the people couldn’t recognize the fake speech and deepfake technology.
PLATFORMS OF DEEPFAKE
There is a Github repository by Aliaksandr Siarohin with the first-order model. It has a torrent and other links to download software called Deepfacelab. All it requires are two videos to be used: source video and destination video. Source video is the video of a user whose face is to be displayed like some celebrity or user himself/herself. Whereas, destination video includes the video of a person whose face is to be replaced like a person giving some speech eg. Elon Musk. After loading these videos in software, both videos are to be resized to 256×256 and transferred to frames. Now face motion should be matched with the voice of the destination video. Erode mask and Blur mask should be used in one frame for creating realistic faces and realistic borders. The frame’s features should be applied to all frames. So, finally, the deepfake video is created. Other software includes FakeApp, DFaker, faceswap-GAN, and Faceswap.
Also Read: Technology making us smart or dumb?
Deepfake algorithms hadn’t produced 100% accuracy in this initial stage. There could be some groups that might illegally and secretly work upon to strengthening this algorithm. Until then, users could spot deepfake by techniques:
- Unnatural eye, body movements, and facial expressions
- Lacking proper emotions
- Unnatural bordering in faces
- Unmatched skin coloring
- Unmatched audio and video
DEEPFAKE & BENEFITS
1. Animation: These platforms could be benefited from creating animated videos where the same frames of emotions could be filtered to provide output with matching facial expressions with body and talk. This would save time and resources if the animation company was previously performing frame by frame or pixel by pixel animation
2. News: In many news channels, the news presenter had gathered a big audience. The listeners watch the news because they are habitual with the recent stories directly from the mouth of the presenters. This is the reason why the presenters aren’t given many leaves. In such a context, deepfake videos could be created with the voice and face of the news presenter even if the news is interpreted by any other person. In small news channels and youtube videos, this method could be used if the presenter doesn’t have time to read articles every day.
3. Personal Transformations: We have seen people performing facial editing in photoshop where a user wants to know how he/she looks with changes in the body or faces, how the person seems after some years, etc. Deepfake could help save time required for photo editing by 2-3 minutes of using the above platforms to view oneself in the video. This could show different avatars of the same user if a person wants to go simple plastic surgeries. It could show how a person looks if he/she wants to have a Korean hairstyle.
4. Art: Leonardo Da Vinci’s best painting Mona Lisa was brought to life through deepfake. Similarly, it could help artists and painters in a simulation of imagined ideas before creation. Artists could well decide whether the paintings should be created or not.
Even with these benefits, numerous crimes are possible to be introduced in the deep fake world of tomorrow. Adding to four of the above challenges, others include financial fraud and blackmail. Deepfake’s huge chances for social technological crimes are sure to attack the internet heavily if the users are much motivated to use it as well as if the developers would keep innovated to provide free and easily online, license-less, and open-source deepfake services in the coming days. Like other newly introduced technologies, it could also take job opportunities from people especially painters and animation artists. This is standing as a big challenge for big internet, website, and data companies of tomorrow to bring new ways of authentication that could beat deepfake. Also, international organizations and government agencies should introduce rules and licenses that would put end to the negative use of this technology.