The creation of realistic audio and video deepfakes has become a pressing issue. The ability to produce highly convincing fake content is no longer confined to high-end production studios but is now accessible to the average user. This democratization of technology has led to significant societal concerns, particularly around cyber threats, misinformation and fraud.
Deepfake technology has been exploited in various malicious ways, such as creating fake videos of CEOs to facilitate phishing scams. Earlier this year, a firm in Hong Kong was tricked into transferring $25 million by attackers using deepfakes. An article from ArsTechnica shared, “The scam featured a digitally recreated version of the company’s chief financial officer, along with other employees, who appeared in a video conference call instructing an employee to transfer funds.”
Deepfakes also pose a serious risk for manipulating public opinion. A growing number of people rely on social media and other online sources for information and are susceptible to convincing deepfakes. With the 2024 elections on the horizon, the potential for deepfakes to impact public perception is alarming.
A study commissioned by McAfee in January of 2024 among 7,000 adults found that nearly two-thirds (64%) of people globally are more concerned about deepfakes today than they were just a year ago, highlighting the growing unease surrounding this technology.
Improving McAfee Deepfake Detector
McAfee has worked closely with Intel to dramatically improve the performance of McAfee Deepfake Detector. On systems where it’s available, the McAfee tool leverages the Neural Processing Unit integrated into Intel Core Ultra processors to enhance the detection and analysis of deepfakes directly on the device. This approach not only boosts performance but also ensures user privacy by processing data locally, eliminating the need to send potentially sensitive information to the cloud.
The McAfee Deepfake Detector, initially showcased as “Project Mockingbird,” utilizes transformer-based Deep Neural Network models that have been expertly trained to detect discrepancies in audio that may indicate a deepfake. This technology was prominently featured at the 2024 CES conference, where McAfee demonstrated its capability to identify manipulated content with remarkable speed and accuracy.
I spoke with Steve Grobman, McAfee’s Chief Technology Officer. He emphasized the importance of this innovation, noting, “In a world where seeing is no longer believing, and AI-generated deepfakes make it harder than ever to tell real from fake, consumers need the latest in AI to beat AI.”
Grobman’s statement underscores the necessity of using advanced technology to counteract the sophisticated methods employed by cybercriminals.
Grobman also stressed that many people believe that the value of artificial intelligence is about building technologies to do something that humans currently do—simply automating routine tasks and expanding scale to streamline productivity. He pointed out that McAfee Deepfake Detector is actually designed to do something humans are unable to do on their own.
“With many of these deepfakes, I can’t tell whether it’s real or fake. But there’s enough signal in the audio that our algorithms can actually pick up on to detect it. Think of it like bionic hearing to be able to tell that it’s likely AI-generated.”
The Power of Collaboration: Intel’s Role
Intel’s collaboration with McAfee on the Deepfake Detector highlights the chipmaker’s commitment to leveraging its technological prowess for the greater good. Carla Rodriguez, Vice President and General Manager of Client SW Ecosystem at Intel, sat down with me to explain the significance of their partnership: “Our collaboration with industry leaders like McAfee underscores our commitment to enabling the technology ecosystem to use AI for good.”
This initiative is part of Intel’s broader AI PC Acceleration Program, which aims to catalyze innovation across the tech industry. Intel has made significant strides in optimizing AI performance on its processors. More than 500 AI models are now optimized to run on Intel Core Ultra processors, showcasing the company’s leadership in driving AI capabilities forward. This optimization enhances various applications, including deepfake detection, by allowing more efficient and faster processing of AI tasks directly on the device.
Enhancing Detection Capabilities and User Privacy
One of the key advantages of the McAfee Deepfake Detector is its ability to perform all processing on the user’s device, leveraging the Intel NPU. This not only speeds up the detection process but also significantly enhances user privacy. According to Grobman, this local processing capability provides substantial benefits to the customer, including a 300% performance improvement over cloud-based processing.
Grobman demonstrated the software for me as well. McAfee Deepfake Detector is capable of running just fine on the standard CPU in traditional PC architecture. However, for systems that have the Intel NPU, McAfee Deepfake Detector can analyze a clip in a fraction of the time it takes to perform the same analysis on a CPU alone.
The local approach also eliminates the latency and potential privacy risks associated with transmitting data to and from the cloud, offering users a more secure and responsive experience. As Rodriguez noted, “When you go from cloud to local, you bring the choice—you can run it on CPU, GPU, or NPU, which each serve a purpose based on the specific application and optimization goals.”
Future Outlook and Challenges
As the technology landscape continues to evolve, the arms race between creating and detecting deepfakes will intensify. McAfee and Intel are at the forefront of this battle, continuously refining their technologies to stay ahead of cybercriminals. The integration of AI into various platforms, especially as we approach critical events like elections, will be crucial in safeguarding public discourse and individual privacy.
Moreover, as these technologies become more mainstream, educating the public about the capabilities and limitations of deepfake detection will be essential. Grobman stressed the importance of this educational aspect, emphasizing, “In today’s digital world, protecting yourself requires both the best technology and a new form of digital street smarts.”
The collaboration between McAfee and Intel serves as a model for how tech companies can work together to address some of the most pressing challenges posed by advancements in AI. By combining their expertise, McAfee and Intel are not only enhancing digital security but are also promoting a more informed and cautious approach to digital content consumption.
