McAfee Project Mockingbird that can detect Deepfake audio demoed

A few days ago, McAfee announced their innovative Project Mockingbird at CES 2024. This is an AI-powered Deepfake Audio Detection technology that uses an advanced AI model from McAfee Labs that is trained to detect AI-generated audio.

At the time of the announcement, the company stated that this technology is over 90% accurate at detecting and exposing maliciously altered audio in videos. Now, it has come up with a demonstration video on YouTube exposing a scam that uses fake Taylor Swift voice generated using AI tech.

The video shown is a scam that promises a fake giveaway and asks the viewer to input some data. Scammers used portions of a legitimate news segment that have been replaced with AI-generated audio to lure victims into enrolling in a fraudulent investment platform.

The red graph shown in the video denotes the output of one of McAfee’s AI models’ predictions, indicating which portions are likely AI-generated. Given that the adversary only manipulated parts of the audio track, the “ground truth” line indicates which portions of the audio are actually “real” vs “AI-generated” and can be used to assess the effectiveness of McAfee’s detection technology.

The Project Mockingbird will help inform users about identifying such AI-generated media content that might be malicious or misleading.