Deepfake or not? McAfee makes it clear by the sound

According to McAfee, AI deepfakes are the next big threat. A new tool from the provider should now be able to check videos automatically and in the background.

Save to Pocket listen Print view
Deepfake lettering in front of cell phone in human hand

(Image: Skorzewiak/shutterstock.com)

3 min. read
By
  • Dr. Udo Seidel

Like any new technology, AI is also misused: OpenAI recently publicly confirmed that cybercriminals have used ChatGPT to develop malware. And fake video and audio material is the starting point for various scams. In the meantime, only a few seconds of a voice recording are needed for a perfect imitation.

McAfee only wants to beat such scams with its own resources. Using software called Deepfake Detector, an AI is supposed to recognize fake video and audio messages almost in real time. The user doesn't have to do anything, not even upload a URL or file to a special verification portal. At the moment, the deepfake detector specializes in English, where the detection rate is 96 percent. Support for other languages is planned for the coming months.

To use the deepfake detector, users must install McAfee Total Protection or LiveSafe. The tool is integrated there and shows statistics on videos that have been checked and classified as dangerous. To get started, this works with the web browser. If you open a website with video material, the deepfake detector checks the associated audio track. If this has been altered with AI, the user is notified. There is also a corresponding note in McAfee Total Protection or LiveSafe.

Why is the video material not checked? Research by McAfee has shown that the sound makes all the difference: a fake video with poor audio quality rarely lures users into the trap –. The average user would rather accept an image with errors and pixels and listen to the associated voices. The deepfake detector also saves computing time and resources.

Nevertheless, the deepfake detector requires computing capacity for its task, but this should not affect the user's actual applications. In technical terms, the software relies on an AI accelerator, also known as a neuromorphic processor. McAfee is conducting a certification process with selected hardware manufacturers for this purpose. Lenovo laptops are already on board, and talks are also taking place with Dell and Asus. Anyone who buys such a laptop can test the deepfake detector free of charge for 30 days. For the following 12 months, the subscription price starts at just under 10 US dollars.

McAfee is also introducing a simple check of suspected deepfake material for private use, called Smart AI Hub. Here, users can find out about the latest developments in the area of deepfakes. And anyone who wants to can help and report fakes they have found themselves.

(mki)

Don't miss any news – follow us on Facebook, LinkedIn or Mastodon.

This article was originally published in German. It was translated with technical assistance and editorially reviewed before publication.