Deepfakes and the Crisis of Visual Trust: How Technology Is Fighting Back

Image Source: Pexels
In 2024, a finance worker from Hong Kong joined a conference call with his CFO and board members. The CFO asked him to transfer $25 million to an overseas account, and so he did.

But after transferring the amount, he discovered that all the people on the call were deepfakes, and he had been scammed, according to CNN. 

This is one of many examples where a deepfake was used to scam people. In the past, people trusted an image or a video as solid evidence for a crime. But with deepfakes, now it’s hard to trust, and video evidence on whether it’s true or fake.

Today, we’ll explore what deepfakes are, how they are impacting us, and what technologies are helping fight back against deepfakes.

What Is Deepfake?

The word deepfake breaks into two words: deep learning and fake. It means creating a piece of video, image, or audio by using advanced artificial intelligence.

This deepfake technology works by using a system called GANs (Generative Adversarial Networks). In the GAN system, two models compete with each other.

One AI tries to create a fake video or audio that looks real. The second AI acts as a critique for the videos, to detect whether the video is real or fake.

These two AIs keep competing and initially improving the fake visuals to a point that it becomes impossible to tell the difference in reliability.

The advanced AI model is trained on thoughts of photos, videos of a person’s face, voice, expressions, and their movements. After they are trained, they can then make that person similar to another video.

This is why judging a visual by the naked eye is no longer enough. That’s why people use an AI image detector to judge if a visual is authentic or has been altered. These tools work by spotting inconsistencies in visuals and verifying whether it’s authentic or has been altered.

How Realistic Are Deepfakes Getting?

What’s frightening about deepfakes is that they are getting more realistic videos day by day. In 2020, the deepfake videos would have obvious flaws in them. This could be something like a blurry face, weird lip movements, or unnatural artifacts around the face.

But by this time, the latest deepfakes have become really convincing to the real one.

Some of the high-quality deepfakes are almost impossible to detect by the naked eye. This is because of the rapid improvement they have had over the past few years.

These AIs don’t just copy the face of a fake person now. They can even replicate micro-expressions, natural eye blinking, head movements, and even skin textures and lighting.

Many deepfakes now pass the Grandma test. This means now the non-technical people can find them completely believable. So it’s continuously getting better day after day.

And as a result, people are doubting to trust real visuals and being confused on what’s fake, and what’s real.

The Crisis of Visual Trust

Deepfakes now have a much bigger problem than just fake videos. They have broken society’s trust in all visual content.

Here are the effects of the issue explained clearly:

Why “Seeing is No Longer Believing”

For over centuries, humans believed in a simple philosophy: if I see it, then it’s real. That has changed now. People can now see a real video of an incident, and can not trust whether it’s an authentic video or made by deepfakes.

The courts used videos as strong evidence for an event, stronger than an eyewitness testimony, or written statements. But deepfakes have broken that basic assumption.

A recent study on humans in detecting high-quality deepfake videos shows that they can now detect accurately around 55-57%. That’s barely better than flipping a coin (50%).

That means, even if you watch a real video today, your eyes can’t be trusted any longer on whether it’s fake or not. The video might be genuine, or a deepfake created, you don’t know now.

The Liar’s Dividend Effect

This one is considered as one of the most overlooked consequences of deepfakes. While most people worry about fake videos fooling the public, the Liar’s Dividend Effect works in the opposite way.

Here’s what that means: when a deepfake becomes highly realistic, someone caught on the camera can claim that “this is a deepfake”. And you can do nothing about it.

Because the technology has become so convincing that the fake becomes believable. So, even when a video is 100% authentic, people will still hesitate, and think. Maybe it’s really fake.

Now because of this deepfakes,

• Politicians can deny their real scandals
• Criminals can dismiss real CCTV footage
• Public speakers can deny any authentic statement they’ve made.

This means the real evidence is now losing its power and the deepfakes are actually helping protect the liars.

That’s why this is called a Liar’s Dividend. Because the liars are getting unexpected benefits from the deepfake technology.

Real Impact on Courts and the Justice System

The deepfake has made a really big impact on the court system. What was thought to be the strongest evidence in courts is now being tested and continuous proof for evidence.

In the past, judges took CCTV footage, mobile recording or surveillance videos as strong evidence for giving statements.

But now, the deepfake videos have become so convincing that the defense lawyers could easily change any videos and simply say that this video could be a deepfake.

Because of this, the prosecutors have to spend a good amount of time and money to prove that the videos are real, not made by deepfake.

In some cases, the authentic video evidence might not be thought out because the court cannot be 100% sure it’s real.

This creates a lot of delay in the trials. What used to be a simple piece of evidence now required technical experts, forensic reports, and analysts for extra verification.

Impact on Politics, Media, and Public Trust

The crisis of visual trust is now also spreading among politics, media and public trust. What used to be a fair election before, the deepfakes became a powerful weapon to manipulate and make fake videos of politicians saying things they never said.

When these videos are proven fake, the initial damage to reputation is irreversible.

The news organizations and journalists are also tracking a serious issue because of deepfake videos. They can no longer confidently show video footage without any thought verifications.

At the bottom level, the people are slowly losing trust from the video deficiency. People are automatically thinking, ” Is this video even real now?”. This doubt has created a climate of confusion where society now finds it hard to believe on the basic facts.

How Technology is Fighting Back

As deepfakes are becoming more dangerous day by day, the technologies are not sitting back. Its rapidly developing power of tools to fight back is restoring the authentic visuals from the deepfake-created ones.

An AI detector has become one of the most promising tools to help with that. These advanced AI systems are trained to detect even the most advanced deepfakes by analyzing only details invisible to the human eye.

These tools look for flaws like inconsistent blinking patterns, unnatural blood flow under the skin, or lighting abnormalities.

Many deepfake image detection tools are also being developed for various users.

These include forensic analysis techniques that test the biological signals, blockchain-based video authentication, and invisible digital watermarking systems. They are used to determine whether the video has been tampered or not.

Final Thoughts

As deepfakes are advancing day by day, it’s becoming harder to trust the visuals as they are becoming more and more realistic. What was once considered unreliable evidence is not being objected to as authentic or not.

However, the good news is that technology is effectively fighting back with its advanced detection tools. With the help of AI image detection and other versatile tools, people can now have valid systems to verify whether the video is authentic or not.