Deepfakes are morphed videos, images, and audio files that are created using deep learning, a form of artificial intelligence. In deepfake videos or images, the face and features of people in the original version are replaced with the faces of others, usually celebrities.
Creating realistic deepfakes is a time and resource-intensive process, but simple, basic deepfakes can already be made using tools that are easily available on the internet. While deepfakes have positive uses in fields such as education, content creation, and journalism, they are mostly used for spreading misinformation, revenge porn, sextortion, and phishing.
Although it’s becoming increasingly harder to recognize them, you can spot a deepfake by any of these telltale signs:
- Unnatural lighting around the video’s subject
- Blurring around the subject’s hair and skin
- A lack of blinking
- A mismatch between the audio and video or a delay in lip-syncing
While it’s unlikely that a deepfake will target you, you can protect yourself by reducing the number of images you share online and installing a VPN and antimalware program to prevent leaks and malware on your device.
In 2021, several videos of celebrities and public personalities saying or doing weird things went viral. For example, in a series of short videos, Tom Cruise narrates his meeting with Mikhail Gorbachev, talks about playing golf, and does a magic trick with a coin. The average viewer is likely to believe that Tom Cruise did, in fact, create these videos. However, that’s not the case.
All of these videos use deepfake technology to replace the creator’s face with Tom Cruise’s. Deepfakes are created using “deep” machine learning, a form of artificial intelligence, to morph someone’s appearance to look and sound like someone else. This article will help you develop a better understanding of deepfakes, how they work, and how they can be identified. Let’s dive in!
What Is a Deepfake?
A deepfake is a fake piece of media, usually a video, that looks real. In gaming and the media, computer-generated images have been used for years already. However, artificial intelligence technology and machine learning have now advanced to a point where it’s possible to accurately manipulate video and audio as well.
Manipulations can extend to seamlessly and convincingly replacing one individual’s face and features with another’s. Take the earlier mentioned deepfake video of Tom Cruise as an example.
These Tom Cruise deepfake videos are extremely convincing, and many users thought they were real. Actually, it’s an actor pretending to be Tom, aided by AI technology.
While this is an exceptionally good example that shows just how far technology has come, most “common” deepfakes are of slightly lesser quality. Even then, however, they can be dangerous. If made well, deepfakes can appear pretty realistic and fool users into believing fake news, disinformation, and misinformation.
How Are Deepfakes Made?
A deepfake is made using a form of artificial intelligence known as deep learning. In simple terms, the process involves feeding the AI algorithm millions of images and videos. The algorithm “learns” from this existing video footage and can create fake videos in which the faces and features of characters are replaced.
Here’s a step-by-step breakdown of how deepfake technology works:
- Collect videos: Deep learning neural networks work best when provided with lots of data. With deepfakes, data consists of video clips of the target. So, anyone wanting to create a deepfake will first need to collect a large number of videos and images of their target. They will also have to identify the video they want to manipulate.
- Analyze videos: The collected videos are fed to an algorithm known as an “encoder.” Its job is to study the videos and identify what the target looks like from various angles and in different environments.
- Map the target onto the video: Once the encoder understands what the target looks like, it can swap the face on the existing video with that of the target. The encoder basically detects similarities between the two faces and determines the best way to swap them without them looking fake.
- Test the deepfake video: This step involves a different kind of AI, known as a generative adversarial network (GAN). The GAN algorithm tests the deepfakes over multiple rounds to find any flaws. Once the GAN no longer detects any flaws, it can be assumed that the deepfake is pretty seamless and accurate.
As you can see, making a deepfake is a pretty intensive process that requires considerable data and computing power. That doesn’t mean it’s rare to run into deepfakes, however.
Apps that let you create deepfakes
Several apps and pieces of software have emerged that let common users create deepfakes as well. Zao App was the first to make deepfake creation widely accessible. Its use has since been restricted in most countries, but people can still access it from some parts of the world, like India.
Zao isn’t the only app that helps you create deepfake videos. Others include:
- Deepfakesweb: This paid browser app lets you merge images and videos.
- DeepFaceLab: This is a more complex program built by Microsoft. Its main purpose is to help students and researchers better understand deepfake productions, but people can use it to create their own deepfake videos, too. It’s a bit harder to use than Deepfakesweb and requires a much larger dataset to create something convincing.
- Wombo: This lip-syncing app lets you stitch faces to music videos. If you want to see it in action, check this clip with Elon Musk singing an unexpected tune.
Most widely available deepfake apps produce fake videos that aren’t very realistic. However, that’s already changing. That’s why it’s important to learn about deepfakes and how they might be used to further someone’s personal agenda.
What Are Deepfakes Used For?
Deepfakes are known primarily for their negative uses, which include revenge porn, phishing, and fake news. However, deepfake technology is increasingly applied in positive ways as well. We examine both kinds of use cases below.
Negative uses of deepfake technology
Deepfakes are notorious for their use as vehicles for false news and cybercrime. Here are some negative ways in which deepfake technology is used.
1. Fake porn
Deepfakes trace their origin to posts by a Reddit user by the name of “Deepfakes.” The user posted morphed pornographic videos with the faces of celebrities and famous individuals on the forum. While his account has been banned, deepfakes have been used overwhelmingly for creating fake porn videos ever since. Research even suggests that 96% of deepfakes are related to porn.
2. Revenge porn
While using deepfakes to make misleading porn videos is extremely worrying, their use for revenge porn is downright scary. Revenge porn is the unauthorized sharing of an individual’s sexually explicit images or videos. Deepfake technology allows people to create explicit images of others, even if these don’t actually exist. It has a particularly bad impact on women, as they are the victims of deepfakes in more than 90% of the cases.
3. Sextortion
Sextortion is a prime example of how deepfakes might be able to do significant real-life damage. Sextortion is a scam in which cybercriminals blackmail victims into sending money, under the threat of releasing compromising videos of said victim. Deepfakes provide a mechanism through which criminals can create actual compromising material to blackmail others.
Even if the victim knows the deepfake video is not real, they may feel they have no choice but to pay, as the videos can be extremely realistic. In January 2021, Avast researched, identified, and blocked over 500,000 cases of sextortion against their clients. Though not all of these work with deepfakes, it does show the scale of the issue.
4. Phishing
The same technology used for sextortion can also be used for phishing. In phishing campaigns, cybercriminals pretend to be someone else to obtain valuable information or incentivize actions.
There are already real-life examples of this. A British CEO was tricked into transferring $240,000 to a fraudster. The CEO believed he was talking to the head of the parent company during a phone call, who asked him to transfer the money urgently. He is believed to have been tricked by a deepfake voice.
5. Fake news and propaganda
In an era where “fake news” seems to contribute to much of the world’s woes, deepfakes have become a propaganda tool. The following deepfake video of Barack Obama shows how deepfakes can be used to manipulate the truth.
There is considerable concern regarding the use of deepfake to influence elections and cause social unrest. As this analysis explains, deepfakes weren’t much of a problem in the 2020 US election. The main reason is that believable deepfakes are still difficult to create. However, in the future, their availability for mass use may increase, and they might truly threaten future elections.
Positive uses of deepfake technology
While most of the news surrounding deepfakes tends to be negative, the technology is increasingly being put to positive and beneficial use. Here are some of the most common positive applications of deepfake technology.
1. Education
Picture this: you’re learning about important historical events, but instead of reading a book or watching an animated clip, you can actually have a historical figure narrate the events for you. This is one of the many applications for deepfakes in education. Overall, deepfakes are expected to make learning more fun for students and will allow teachers to create more innovative lessons.
2. Film and media
CGI and VFX have been standard aspects of films and media for several years now. They add a visual layer that helps capture the audience’s imagination. Deepfakes can make these techniques accessible to a wide range of filmmakers and producers. In time, we may see influencers and social media personalities use deepfakes to create new engaging content.
Deepfakes can also be used to recreate and animate characters played by actors who aren’t actually on set. In a recent Star Wars movie, The Book of Boba Fett, deepfake technology was used to recreate a young Luke Skywalker using footage from earlier movies.
3. Content personalization
Since deepfakes allow video and voice to be manipulated, content can be customized and personalized for specific users. David Beckham’s campaign with Malaria No More is a great example. The initial video was shot in English, but deepfakes were created in several other languages. These were far more believable than average translations as Beckham’s lips were synced to the audio and the translation was added in his voice.
Moving forward, we can expect to see deepfakes used as a tool to create content that caters specifically to individual users or clients in different languages and regions.
4. Anonymity
Deepfake technology and synthetic media can also play an important role in helping journalists, activists and researchers maintain their anonymity. These individuals are often the target of oppressive regimes, so concealing their identity is crucial to their work.
The documentary “Welcome to Chechnya” provides a great example of how deepfakes can be used to protect the identity of potentially vulnerable individuals. It uses deepfake technology to map the faces of volunteers onto the victim’s subject. Since the actual victims are members of the persecuted LGBTQ community, this use of deepfakes helps keep them safe.
How to Identify a Deepfake
Since deepfakes are increasingly common now, it’s vital that we learn how to spot them. While well-made deepfakes can be hard to recognize, more rudimentary ones have some typical deficiencies you can look out for.
- Blurring around the skin and hair is common in deepfakes that aren’t made well. This is especially likely to occur when the individual in the deepfake video changes from front profile to side profile or vice versa.
- The lighting surrounding the individual can look unnatural, since the deep learning algorithm retains the lighting of the original video.
- The lack of blinking by the individual in the deepfake is a telltale sign. However, since research that highlighted the blinking issue was published, creators of such videos have improved the technology. Blinking is now common in most deepfakes.
- A mismatch between the audio and video is another sign to look out for. This is particularly common in low-quality deepfakes where the cybercriminal doesn’t make the effort to properly sync up the two.
Moving forward, artificial intelligence and deep learning tools that help identify deepfakes are going to become more common. Leading companies, including Facebook, Microsoft, and Google, have already been developing tools and datasets that help identify deepfakes.
In addition to the above pointers and tools, the most critical aspect of spotting a deepfake is critically analyzing the content you’re viewing online. Deepfakes are certainly a problem, but there is a lot of other low-end false and misleading content out there. Hence, it’s advisable that you look at everything in a larger context and determine facts for yourself instead of relying on the media being fed to you.
How to Protect Yourself Against Deepfakes
Creating a realistic deepfake requires an investment of time and capital. Unless there’s a strong incentive, it’s unlikely that a cybercriminal will make a deepfake video featuring you if you’re an average internet user. Regardless, you can take steps to limit the possibility of a deepfake being created in the first place.
1. Limit the number of pictures and videos you share
Sharing images online isn’t a bad thing. However, if you take many pictures of yourself and make them publicly available, you’re an easier target for cybercriminals employing deepfake technology. Remember that the most realistic deepfake videos can only be created with a large data set of someone’s pictures and videos.
Of course, it’s almost impossible in today’s day and age to completely stop sharing photos online, but you can do so in a more privacy-friendly manner. Before posting something, you can always ask yourself whether sharing that picture of yourself is really needed. Keep the number of shared pictures low, if possible. You might also use temporary forms of photo sharing, such as Stories or Snaps, which vanish after 24 hours. If you’re good with photo editing software, you can insert a digital watermark in your content to prevent its use in deepfakes.
2. Keep your social media profiles private
Most social media platforms allow you to keep your profiles private. We recommend enabling this setting as it limits the scope of hackers accessing your photographs. Additionally, you can choose to limit the people who can tag you in photos or interact with you on social media. This can also help prevent your photos from becoming publicly viewable.
We’ve written detailed guides on optimizing privacy settings for a number of social media websites, including LinkedIn, Facebook, and Snapchat.
3. Secure your online activity with a VPN
You should also make sure your devices are secured, so your data (including images and videos) doesn’t get leaked as easily. This can be done by using a VPN (Virtual Private Network), which helps you stay anonymous online.
While using a VPN doesn’t directly protect your photos or videos, it does prevent your browsing activity from being prayed upon by others. Hackers often identify targets through their browsing history or by intercepting their internet connection. Using a VPN is more of an overall security measure that can help protect you against cybercriminals and curious eyes.
We recommend getting a subscription to NordVPN, as it’s known for its stellar security, privacy features, and excellent speeds.
4. Protect your devices with an antivirus scanner
Most of us store personal photos and videos on our devices. If these were to be compromised by a hacker, they could potentially use them to create deepfakes. Hence, it’s important to ensure that your device is free of malware, including trojans and viruses.
The best way to keep your device malware-free is by using a reliable and trusted anti-malware scanner, like Norton 360. Norton doesn’t just scan your device for threats, but also blocks external threats from the internet.
5. Contact the authorities
If a deepfake featuring you was created and is circulating online, you can report it immediately to the social platform it was shared on. Deepfakes violate the community guidelines of most platforms, so the content should be taken down swiftly.
Besides that, you can also contact local authorities to make a formal complaint. There are various helplines run by non-profit organizations that can help you out here. For example, the Cyber Civil Rights Initiative runs a helpline where US citizens can report image-based sexual abuse and seek guidance and counseling.
A Deepfake Futurescape
Like many other technologies, deepfakes have good and bad uses. They are currently known for their negative impact, which includes creating fake videos and spreading misinformation. The disastrous consequences of this developing technology are already being discussed online and even in TV shows like The Capture. However, deepfakes can be used positively as well, and such uses may become more prevalent in the near future.
Since deepfakes can be used for propaganda and cybercrime, including sextortion and phishing, it’s important to learn how to spot fake videos. Some telltale signs include distorted lighting, blurry edges, a lack of blinking, and a mismatch between the audio and video.
Though the chances are small that someone will make a fake video featuring your likeness, you can take measures to protect yourself and your online presence. Limiting the amount of video footage you share of yourself publicly already helps a lot, as well as protecting your device from outside attacks with good antivirus software and a solid VPN.
Want to learn more about online scams and the creation of “fake” online media? You can read more in the following articles:
- What is Social Engineering?
- How Invasive are Face Transforming Apps?
- Top Instagram Scams and How to Avoid Them
- Create an Avatar with AI: Top 8 Safest AI Portrait Generators
Deepfakes are becoming more readily available and easier to spread. Do you have a question about this new way of creating “fake” media? Check this FAQ section for our answers to the most frequently asked questions about deepfakes.
A deepfake video is a piece of media created with deepfake technology that incorporates video and audio elements. Deepfake technology uses AI and machine learning algorithms to modify existing videos and add new faces or voices over the original file. Many popular deepfake videos feature a celebrity saying or doing something ridiculous or outrageous. You can find out more in our article about deepfakes.
The term “deepfake” was first used by a Reddit user going by that name back in 2017. This user created fake porn videos featuring various celebrities. However, this is not necessarily the first deepfake ever created, as AI is known to have been used for video editing well before that time. While no exact moment can be defined, it’s safe to assume that the first deepfake was created less than ten years ago.
Making realistic and accurate deepfake videos involves training AI algorithms to detect and replace faces and features in existing footage. It’s a computer-intensive process and requires a lot of data. However, less convincing deepfakes can be made using deepfake apps that are readily available on the internet.
