Exploring the Accuracy of McAfee's New Artificial Intelligence System for Identifying Fake Audio

Can McAfee’s New Artificial Intelligence-Powered DeepFake Audio Recognizer Stand the Test of Time? Exploring Its Dependability

Cemile Bingol/Getty Images

Artificial intelligence (AI)-generated deepfakes are all over the internet, fueling everything from scams to political misinformation. Cybersecurity provider McAfee is now the latest entity to attempt a solution – but it’s not for everyone, and it may not be a catch-all.

On Wednesday, cybersecurity provider McAfee launched its Deepfake Detector, an AI-powered tool that scans videos for signs of AI-generated audio. According to the release, once a user opts in, the detector automatically monitors what’s playing on a PC and alerts the user “within seconds” if it finds AI-altered audio – all without the use of third-party software or the need for user-initiated manual checks.

Also: AI phone scams sound scary real. Do these 5 things to protect yourself and your family

Altered audio could indicate a deepfake scam, including those that clone the voices of loved ones to trick people into paying fake ransoms or impersonate public figures to spread false information. McAfee’s release notes that the proliferation of AI has enabled cybercriminals to create “more convincing, personalized, AI-generated scams at scale.”

“While not all AI content is created with malicious intent, the ability to know if a video is real or fake helps consumers make smart and well-informed decisions,” McAfee’s announcement states.

“Deepfake Detector only analyzes audio; more specifically, it scans audio on videos playing on the browser and does not scan things that involve digital rights management, like the content you’d find on a video streaming service like Netflix or Disney+,” a McAfee spokesperson told ZDNET.

Also: Most people worry about deepfakes - and overestimate their ability to spot them

The detector does not monitor any video or audio playing outside of the PC. “Our threat research found that most deepfake videos used AI-manipulated or AI-generated audio as the primary way to mislead viewers. The visual elements of the videos are typically tightly cut, genuine footage to distract the viewers from the suspicious message,” the spokesperson continued.

McAfee created Deepfake Detector in partnership with Lenovo specifically for its Copilot+ PCs , meaning it is only available on those devices, and will eventually cost at least $10 per year.

Newsletters

ZDNET Tech Today

ZDNET’s Tech Today newsletter is a daily briefing of the newest, most talked about stories, five days a week.

See all

Using McAfee’s Neural Processing Unit (NPU), the company’s AI detection models run inference, or the end-to-end content identification process, entirely on the PC. This kind of on-device AI reinforces user privacy by keeping personal data out of the cloud – an approach Apple is taking with Apple Intelligence , rolling out to devices this fall .

Of course, the detector must surveil a user’s laptop to work – but McAfee assures this won’t infringe on privacy. “McAfee does not collect or record a user’s audio in any way, and the user is always in control and can turn audio detection on or off as desired,” the company explained in the release.

McAfee also explained that the benefits of running Deepfake Detector on-device include faster processing speed and improved battery life.

Also: 80% of people think deepfakes will impact elections. Here are three ways you can prepare

But does AI audio detection really work?

“No more wondering, is this an actual celebrity giveaway, is this investment offer legitimate, did this politician actually say these words?” said Roma Majumder, McAfee’s senior vice president of product, in the release.

That’s a high expectation to set. Putting this kind of technology – however early-stage – directly onto devices better prepares consumers for a future full of AI-generated content. However, as ZDNET’s own reporting has found, the accuracy and efficacy of detection technology have been unreliable, especially considering the pace at which generative AI is evolving.

“The technology uses a model trained across nearly 200,000 samples and gives consumers advanced AI detection, with a 96% accuracy rate,” the McAfee spokesperson explained of the detector.

Also: More political deepfakes exist than you think, according to this AI expert

Alongside Deepfake Detector, McAfee also released the Smart AI Hub , an educational cybersecurity resource for consumers aiming to “build awareness of deepfakes and AI-driven scams,” according to the announcement. Users can submit suspicious-looking videos to the hub for McAfee to analyze. The company says it plans to share the resulting insights as a resource with the public.

The Deepfake Detector is available today in English for all new Lenovo Copilot+ PCs ordered from Lenovo’s website and select retailers in the US, Australia and the UK. Once they’ve purchased the PC, a customer can use the detector for free for 30 days, after which pricing begins at $10 for the first year.

The feature will roll out to more countries later this year.