What Is Deepfake and How Does It Work?

Over the last few decades, computers have advanced dramatically and become increasingly adept at simulating reality, owing primarily to the rapid development of AI. Artificial Intelligence has numerous applications in various fields, such as natural language processing, speech recognition, machine vision, etc. However, not all AI products benefit society because an attacker may employ some of these technologies in their illegal activities, and deepfake is one such technology.

So. what is deepfake, and how does it work? Before answering these questions, let’s first go through the list of topics that we will be covering in this blog:

• What is deepfake?
• How is deepfake created?
• How does deepfake work?
• Real-life deepfake cases in India
• Is deepfake that dangerous?
• How to detect deepfake?
• Conclusion

What is deepfake? 

Before we get to the definition of deepfake, let’s first try to understand what the word “deepfake” means. The “deep” part of the word “deepfake” refers to deep learning, a technique of teaching machines to think like human minds. The term “fake” emphasizes the misleading nature of deepfake media.

Deepfake can be an image, video, and audio that appear natural but are artificial intelligence-manipulated fabrications. Attackers commonly use deep learning, machine learning, and artificial intelligence to create fake content, such as superimposing a celebrity’s face onto another person’s body in order to deceive viewers. Deepfakes are different from fake images as fake images are created by humans using tools like Photoshop. Whereas deepfakes are mainly created using artificial intelligence. Thus, making it harder to protect yourself against identity theft.

It’s not like deepfake assists in carrying out only bad or illegal things. Consider deepfake videos, which we frequently use in memes, social media filters, and face-swapping apps. Almost everyone, especially people in their twenties and thirties, has used them on social media platforms such as Instagram or Snapchat. So, we can also use them for entertainment purposes.

You can also explore: Types Of Hackers To Be Aware Of In 2022

How is deepfake created?

Deepfakes are created by feeding many images of the target to deep learning computer networks known as VAE (variational auto-encoders). The objective is to teach the VAEs to collect a variety of specialties, such as lighting conditions, positions, and facial reactions. These specialties assist the AI in determining which visual elements, such as expressions or shadows, are distinctive and which can be removed or changed.

Let’s take an example to understand this complete process:

Suppose there are two celebrities. The first celebrity is Alison Brie, and the other is Jim Carrey. Now the attacker will provide artificial intelligence with input images of the source (Alison Brie) and the target of the deepfake (Jim Carrey). After receiving the input images, the AI begins to create the output images while assessing which subtleties of expression are distinctive and crucial to the deepfake. To be believable, the AI will maintain personality nuances and facial twitches that naturally exist and are distinctive to the target individual (Jim Carrey).

After that, the AI will merge the input and output images to perform the face swap. Using Jim Carrey’s face, the encoder will re-create Alison Brie’s movement patterns and facial reactions.

Note: If the attacker desires to produce a deepfake video, the photos must be merged frame by frame in order for the deepfake clip to be genuinely believable.

You can also explore: What Is Session Hijacking Attack and How To Prevent It?

How does deepfake work?

As stated in the preceding example, AI generates content based on the data input. If a program has to create a new face or replace a portion of a person’s face, you should first train that program. In order to accomplish that, the program is fed massive amounts of data, which it then uses to learn and generate its new data.

You can also explore: What Is Malware? Types & How to Prevent?

Real-life deepfake cases in India

Deepfake has thick roots in India, where it is primarily in use in politics, film, pornography, defamation, and other areas. From deepfake depicting Mr. Manoj Tiwari insulting the Delhi state government to the generation of pornographic content, such as that of Ms. Rana Ayyub, it is clear that deepfake can cause real damage by hurting people’s feelings, opinions, and perceptions.

Let’s look at some of the famous deepfake cases that happened in India:

Case1: 

In Ahmedabad, a 46-year-old businessman became entangled in a sextortion scheme. This happened after accepting a friend request from a woman on social media. They exchanged numbers, and the woman made a video call within two days.

According to the victim, he connected with the women on a mere 15-second video call. After around 15 minutes, he received an extortion call demanding ₹ 2 lakh or else they would make the video go viral. The victim informed the cyber-cell about the entire incident. The police investigated the situation. They discovered that the woman in the video clip was a deepfake created using advanced technology such as AI and ML.

Case 2: 

A deepfake of Manoj Tiwari went viral on WhatsApp in India ahead of the Delhi legislative assembly elections. Manoj Tiwari speaks in English in the original video, criticizing his political opposition Arvind Kejriwal. In the video, we can see him inspiring citizens to vote for the BJP. The second video was altered with deepfake technology so that his mouth moves as he speaks in Haryanvi.

Is deepfake that dangerous? 

Deepfake technology is evolving at a breakneck pace. And it won’t be long before this technology advances to the point where it’s impossible to tell fake videos from real ones. When this occurs, the consequences can be disastrous, especially for public figures and celebrities.

Malicious deepfakes can jeopardize and even ruin many careers and lives. People with malicious intent could use these to impersonate others and manipulate their friends, families, and coworkers. They can even begin global conflicts and wars using fake videos of global leaders.

Students from all over the world participated in a survey. According to the survey, people remember false information more than real news. According to the results of the survey, voters may establish false beliefs after viewing fabricated news stories, particularly if those stories coincide with their political beliefs. Thus, in my opinion, “Yes,” the deepfake technology is hazardous.

You can also explore: What Is Eavesdropping & How To Prevent It?

How to detect deepfake? 

At the moment, it may be possible to detect poorly generated deepfakes with the naked eye. However, as technology advances, it will soon be impossible to tell whether a video is genuine or not. In the beginning, for example, an unusual blink of the eyes was a good indicator of a deepfake. However, systems have learned to imitate eye blinking over time.

This is because the technologies that detect deepfakes are also based on artificial intelligence. This is accomplished by employing algorithms similar to those used to generate deepfakes. So, whenever you encounter a situation in which you are unsure whether the displayed video, image, etc., is genuine or not, use critical reasoning and ask yourself questions, such as:

• Who or what is the originator?
• When and where was the video shot?
• Who will benefit from watching this video?
• Who and why is this video being distributed?
• Is the person in the video saying something unexpected?
• Is the video or image promoting someone else’s agenda?

When compared to a video, it is much more challenging to detect an image created with a deepfake. Using deepfakes to create a video requires much knowledge and expertise. As a result, it is not difficult to detect until and unless a deepfake video is created flawlessly. Let’s look at some flaws that can help you identify a deepfake image or video.

• Jerky motions
• Lighting issues
• Unreal skin tone
• Skin tone variations
• Extra pixels in the frame
• Speech and lip movement are not synchronized
• The person’s face is blurrier than the background

You can also explore: What Is Penetration Testing and How Does It Work?

Conclusion

Deepfakes are a relatively new and promising technology. However, many experts believe that deepfakes will become far more sophisticated as technology advances. As a result, the public may face more severe threats, such as election tampering, political conflict, and extra law-breaking.