Combating the Rise of Deepfakes: A Growing Threat to Security
As AI Technology Evolves, So Do Scams and Security Challenges
Tech firms are currently engaged in a battle against the rising threat of deepfakes—realistic voices and videos generated by artificial intelligence that can deceive individuals and organizations alike. With advancements in generative artificial intelligence (GenAI), scams utilizing these technologies have become alarmingly sophisticated.
A Distressing Reality for Victims
Consider the harrowing experience shared by Debby Bodkin, whose 93-year-old mother received a fraudulent phone call. A cloned voice, mimicking a family member, claimed: “It’s me, mom… I’ve had an accident.”
When asked about her location, the impersonator cited a hospital, furthering the ruse. Thankfully, it was a granddaughter who answered, promptly calling Bodkin at work to confirm her mother’s well-being.
The Alarming Frequency of Scams
“It’s not the first time scammers have called grandma,” Bodkin informed AFP, pointing out that these incidents are becoming a daily occurrence.
Techniques Behind Deepfake Scams
Such deepfake phone scams often aim to manipulate victims into sending money for fictitious emergencies or medical care. Beyond individual targets, deepfakes are increasingly being used on social networks, impersonating celebrities and public figures for malicious purposes, including disinformation campaigns.
Criminal Exploitations of GenAI
In a striking example from Hong Kong, police revealed that an employee of a multinational firm was duped into wiring HK$200 million (approximately US$26 million) after being tricked by AI-generated avatars of his colleagues during a videoconference.
The Difficulty of Detection
A recent study conducted by identification start-up iBoom found that only a minuscule tenth of one percent of people in the U.S. and U.K. can accurately identify a deepfake image or video. This alarming statistic underscores the challenge of distinguishing authentic content from AI-generated fake.
Advancements in Deepfake Creation
Ten years ago, creating a synthetic voice required 20 hours of recorded material. Today, thanks to breakthroughs in GenAI, that process has been reduced to a mere five seconds, according to Vijay Balasubramaniyan, CEO of Pindrop Security.
Detecting Deepfakes in Real-Time
In response to these threats, companies like Intel have developed tools to detect AI-generated audio and video in real time. Intel’s “FakeCatcher” technology identifies color changes in facial blood vessels to differentiate between genuine and fabricated imagery.
Similarly, Pindrop analyzes each second of audio, comparing it to known human voice characteristics to unveil possible fakes.
The Importance of Staying Updated
“You have to keep up with the times,” asserts Nicos Vekiarides, chief of the Attestiv platform specializing in authenticating digital content. He emphasizes the technological evolution, noting that early deepfakes displayed obvious flaws—such as people with six fingers—making them easier to spot.
The Rising Global Threat
Balasubramaniyan describes the growing issue of deepfakes as a “global cybersecurity threat,” as any company can suffer reputational damage from fake content or fall victim to sophisticated fraud.
Adapting to the New Normal
The shift toward remote work has created more opportunities for scammers to impersonate employees, making the need for robust detection technologies even more urgent.
Consumer Solutions in the Fight Against Deepfakes
Ahead of the curve, Chinese tech company Honor launched its Magic7 smartphone, which features a built-in deepfake detector powered by AI. Meanwhile, the British start-up Surf Security debuted a web browser designed to flag synthetic voice or video content for businesses.
The Future of Deepfake Detection
According to Siwei Lyu, a computer science professor at the State University of New York at Buffalo, deepfakes will eventually become comparable to spam—an internet menace that society will learn to manage. He predicts that detection algorithms will function similarly to email spam filters.
“We’re not there yet,” Lyu cautioned, indicating the ongoing need for innovative solutions.
Frequently Asked Questions
- What are deepfakes?
- Deepfakes are realistic synthetic media, including audio and videos, generated by artificial intelligence that can impersonate real people.
- How are deepfakes being used by scammers?
- Scammers use deepfakes to create convincing impersonations of individuals to trick victims into providing money or personal information.
- What technologies are being developed to combat deepfakes?
- Firms like Intel and Pindrop Security are developing tools and technologies that can detect AI-generated content in real time.
- Why is it difficult to identify deepfakes?
- Most people lack the training or tools to discern deepfakes, with only a small percentage able to accurately identify them, leading to widespread deception.
- What future developments are expected in deepfake detection?
- Experts predict that as deepfakes become more prevalent, there will be a greater need for detection algorithms akin to email spam filters to help identify fake content.
