FTC Names Winners in Voice Deepfake Challenge

The Federal Trade Commission (FTC) has announced the winners of its Voice Cloning Challenge, aimed at fostering innovative solutions to combat the misuse of AI-enabled voice cloning technology in fraud and other harmful activities.

The challenge recognized four submissions that propose diverse approaches to enhance consumer protection against deceptive voice cloning. The winners include AI Detect by OmniSpeech, focusing on using AI to distinguish between real and synthetic voices; DeFake by Dr. Ning Zhang, which suggests adding distortions to voice samples to prevent accurate cloning; and OriginStory by the team from OriginStory, aiming to authenticate the human origin of voice recordings with existing device sensors.

Pindrop Security, a larger organization that is well known in the biometrics industry, received a Recognition Award for its real-time detection technology of voice clones and audio deepfakes. Its technology analyzes audio in two-second segments to flag potential deepfakes, enhancing the ability to protect against fraudulent activities utilizing cloned voices. Although Pindrop, as a large organization, was not eligible for monetary prizes, its contribution was acknowledged through the aforementioned Recognition Award.

The Voice Cloning Challenge was designed to spur innovation in safeguarding consumers from the potential misuse of AI-enabled voice deepfake technology. Recognizing the dual-edged nature of voice cloning—which can offer significant benefits like aiding those with speech impairments, but also poses risks such as enabling scammers to commit fraud—the FTC sought to engage the ingenuity of the American public to develop protective measures.

Participants were encouraged to propose solutions that could detect, differentiate, or authenticate voices to prevent deceptive practices. The challenge attracted submissions from individuals and organizations proposing a range of technologies, from AI-based detection systems to unique watermarking techniques that secure voice samples against unauthorized cloning.

With prizes totaling $35,000 for the top submissions from individuals and small organizations, the challenge aimed to highlight and foster innovative approaches to ensure the ethical use of voice cloning technology while maintaining consumer trust and security. That prize money will be shared by the smaller teams behind the AI Detect, DeFake, and OriginStory systems.

Sources: FTC, Pindrop

–

April 10, 2024 – by Alex Perala and the FindBiometrics Editorial Team