In recent years, more and more concern has been expressed about the use of deepfake technology and how it can be used to deceive people and alter public opinion. According to deepfake expert Dharmender Salian, “The rapid development of deepfake technology is briskly transforming how we perceive truth in the digital era. It profoundly impacts everything from entertainment to misinformation, posing both a threat and a powerful tool.”
The Dangers of Deepfake Technology
Scammers often use deepfake technology for financial gain. In February 2024, a Hong Kong company was scammed out of $25 million dollars through deepfake recreations of a coworker’s colleagues. Similarly, scammers have used this technology to recreate the voices of a victim’s family members and con older adults such as grandparents out of their money in a fraud known as the “grandparent scam.”
How Deepfake Technology Works
To produce a deepfake voice, image, or video, scammers commonly use artificial intelligence and machine learning. Scammers grab a sample of an audio clip or an image, then use computer technology to create a verbal message or an altered image.
According to WebWise, scammers use a machine learning technique called a generative adversarial network (GAN) for deepfake videos. This technology uses facial mapping to divide a person’s face into points, and GAN can also be utilized to generate new audio clips from an existing audio clip (such as a snippet of a person’s voice) or new text from existing text.
The Guardian notes, “It takes a few steps to make a face-swap video. First, you run thousands of face shots of the two people through an AI algorithm called an encoder. The encoder finds and learns similarities between the two faces, and reduces them to their shared common features, compressing the images in the process.
“A second AI algorithm called a decoder is then taught to recover the faces from the compressed images. Because the faces are different, you train one decoder to recover the first person’s face, and another decoder to recover the second person’s face.
“To perform the face swap, you simply feed encoded images into the “wrong” decoder. For example, a compressed image of person A’s face is fed into the decoder trained on person B. The decoder then reconstructs the face of person B with the expressions and orientation of face A. For a convincing video, this has to be done on every frame.”
Dr. Raed Omar Sbeit, a deepfake expert and adjunct professor for the University, observes: “As AI models become more sophisticated, the challenge will shift from detection to mitigation, requiring stronger authentication methods and digital literacy to combat misinformation. Just as buyers verify a car’s history with a VIN check before making a purchase, people should fact-check videos and images using reverse image searches, metadata analysis, and trusted news sources to ensure authenticity before believing or sharing them.”
Detecting Deepfakes
Because deepfake technology is so convincing, it is difficult – but not impossible – to detect. One method relies on applying critical thinking and common sense.
Scam artists often use highly sophisticated technology and convey a sense of urgency to pressure you into acting quickly. But using logic can be useful in these situations.
For instance, imagine that you saw a deepfake video where a person took an action that was completely out of character or physically impossible. Wouldn’t you stop to think if it was real or not?
Similarly, imagine that you got a phone call from a scammer posing as an “arresting officer.” Through cloning your teenager’s phone number, the “arresting officer” said that your child was under arrest and urgently needed bail money.
However, that officer then asked you to send the funds as gift cards or money orders rather than coming to a police station. Shouldn’t that request raise red flags and motivate you to check on your child through another source?
Another method of detecting deepfakes includes paying sharp attention to a video’s details. Some clues to a deepfake video include:
- Face/body proportions that don’t match up
- Inconsistent or non-existent eye blinking
- The length of the video (fake videos are usually only seconds in length due to the time it takes to train an algorithm)
- A video is missing its audio track, or the audio doesn’t match up with a speaker’s lip movements
- Blurred areas inside the mouth when someone speaks (deepfake technology has a hard time reproducing the inside of a mouth when someone talks)
- Looking at small details, such as blurry shadows, overly wrinkled or overly smooth skin, and unnatural lip colors
Deepfakes Aren’t Illegal, But That Situation Is Changing
Deepfake technology is concerning because it can be used to spread misinformation, create pornographic images without a victim’s consent, scam individuals or businesses, or conduct cyberattacks. Currently, there isn’t much legislation against the use of deepfake technology in the U.S., but that situation is changing.
There are federal and state laws in development to protect against the use of digital forgeries such as deepfakes. Although some creators may claim that creating deepfakes could be considered free speech under the First Amendment, the First Amendment has some exceptions when it comes to using visual or audible content for fraud, harassment, or misinformation.
Artificial intelligence and machine learning have many wonderful applications for business, education, healthcare, and many other fields. But at the same time, society will need to protect against the misuse of artificial intelligence and machine learning to protect people from harm.
Comments are closed.