Deepfake: Everything You Need to Know to Protect Yourself

Separating fact from fiction while browsing the web has only become more challenging during the past decade. The increased reliance on electronic communication and virtual collaboration tools has made people more susceptible to not question the information they are consuming.

With the overall cost of fake news totaling more than $78 billion globally every year, recognizing the signs has become crucial for companies’ profitability. The development of technology such as deepfake has increased fabricated information’s production, ballooning its cost.

This article will define and examine deepfakes, highlight their impact on cyber security, and showcase how you can protect yourself from being a deepfake target.

What is a Deepfake? The History Behind this Innovative Technology

In simple terms, deepfakes are falsified images, videos, and audio recordings created by using computers and software. Videos and images are blended to create synthetic content that appears authentic. The term is a blend of the words deep learning and fake.

It’s hard to establish when this technology was invented since it derives from photos and digital video manipulation technology that dates to the 1990s. However, the first successful deepfake attempt was created and published on Reddit in 2017. The user posted maliciously altered explicit videos, placing celebrities’ faces over the actors.

How Does a Deepfake Work? Exposing its Hidden Face

Creating a believable deepfake requires the use of artificial intelligence (AI). The two most popular techniques use deep learning or a generative adversarial network (GAN).

In the first method, thousands of pictures of two people are put in an encoder. The AI algorithm then compares all the material provided to find similarities and compresses the images. Next, decoders, one for each person, retrieve the faces from the compressed images. Person A’s features are put on the face of Person B and vice versa, creating convincing synthetic content.

In the GAN method, two machine learning (ML) work against one another. The first algorithm, the generator, uses data to produce synthetic images. Then, the second algorithm, known as the discriminator, combines the synthetic images produced by the generator and real images. Repeating the process numerous times allows the ML to create realistic faces of random individuals. This technique demands the most time and is the most efficient method to create deepfake images.

In the case of deepfake audio, all original material is run through an AI algorithm that clones the person’s voice, tone, and mannerisms. Speeches, presentations, interviews, phone calls, and corporate videos free from interruptions or ambient noise can be used to create deepfake audio.

How Easy is it to Create a Deepfake?

The access to software enabling the creation of deepfake has drastically increased in the past couple of years. As a result, people with no photoshopping or technological skills can now create such content.

Mobile phone apps such as Zao allow individuals to place their faces on actors in pre-selected movies and TV show scenes. The process requires as little as one photo of the user. The app’s algorithm then swaps the person’s face with the actor’s, creating credible results.

Software and desktop apps such as FakeApp are also factors in this issue. Even though they are relatively complex to use, the programs provide instructions, making them easy to use for anyone with moderate technical skills.

It is also important to note that a new Chinese AI algorithm can convincingly clone a voice using a short audio recording of fewer than 4 seconds. Meanwhile, enough important voice traits are decoded and captured. Completing the same task required 30 minutes of audio content and multiple audio clips a year prior.

Nevertheless, only experts can develop high-quality deepfake. High-end computers are required to remove imperfections and visual defects. This process can take up to a couple of weeks, depending on the format and the software used.

What Are the Impacts of a Deepfake on Cyber Security?

Although many attest to the beneficial aspects of deepfakes, such as for entertainment purposes, the technology represents a threat to cyber security. Video and audio manipulation facilitate vishing and spoofing by making identity theft more believable.

Deepfake cyber attacks can target anyone. Easy access to deepfake apps, software, or even websites makes everyone susceptible to being a victim. Because of social media, it is also possible for your identity to be used in a deepfake attack. Photos and videos of yourself posted online can be retrieved and placed in the software.

In the context of a corporation, deepfake usage to modify audio content aids scammers to impersonate employees, suppliers, or executives of a specific company during phone calls. The increasing presence of companies on social media has an immense effect on their vulnerability to such a threat.

A famous case of fraud using deepfake made the news in early 2020. After having a phone conversation with who he believed was a company’s director, a bank manager in Hong Kong authorized the transfer of $35 million. The bank manager previously chatted with the director, but that did not prevent the manager from getting tricked.

Four tactics to Identify and Fight Against Deepfake

To protect yourself from deepfake, four main tactics can be put in place.

1. Do a visual analysis of the content you are viewing. This includes analyzing the way the person in the video blinks and the quality of the image.
2. Stay alert. It is crucial to develop a critical mindset and verify the content’s source.
3. Use AI to detect any anomaly that can be harder for the human eye to catch.
4. Start phone conversations with a secret password or question to ensure the speaker’s identity.

For additional information on ways to protect yourself in case of a deepfake phone call, check out our Don’t be a Victim of CEO Fraud blog.

Recap

The increase in the use of deepfake and the lack of legislation around the subject are cybersecurity threats. The advances in artificial intelligence have facilitated the production of persuasive deepfake. Therefore, it is crucial to be aware, stay vigilant, and practice cyber security. This includes being careful with what you publish online and not believing everything you come across.