Deepfake Detection in GenAI Voices - “Laboratory” Benchmark and Accuracy Report

Presented by Deep Media AI (See our other research in Deepfake Detection here)

https://www.loom.com/share/888127ff903f437bb557b598fcc2614b

Published: 2024 - 07 - 01

Updated: 2024 - 09 - 01

Dataset ID: 240705-FB1CD9

Unveiling The World’s Largest Public Benchmark for High-Fidelity GenAI Voice Detection

Download This Dataset

Fill out this form with your information and someone from the Deep Media research term will grant you access: https://docs.google.com/forms/d/e/1FAIpQLSeTupVdPJDJM6hAEYDCD5ndIu1lvqZhc_xR_woC5hike532hw/viewform?usp=sf_link

Our Academic Research Partners

image (48).png

image (45).png

image (44).png

image (47).png

image (46).png

image (42).png

image (43).png

Abstract

Now that Generative AI can produce hyper-realistic voice content with unprecedented speed and fidelity, the threat of audio-based AI Disinformation has rapidly evolved from a theoretical concern to an urgent societal challenge. As the gold standard in Deepfake Detection and AI Disinformation Security, Deep Media stands at the forefront of combating this escalating digital crisis.

Generative AI Model/Service Breakdown

Generative AI Model/Service Breakdown

We are proud to introduce our groundbreaking Deepfake Detection In-The-Wild Benchmark dataset for GenAI Voice detection, a critical resource for our government, enterprise, and academic research partners:

Diverse array of cutting-edge voice generation algorithms, encompassing 15+ state-of-the-art systems
200,000+ meticulously curated audio samples, representing the largest and most comprehensive collection of high-quality AI-generated voice content
Samples across 8,000+ different speakers diverse in age, race, gender, and emotional expression. Many of these speakers are present in both the GenAI and Real samples. The number of speakers and inclusion of many speaker identities across both GenAI and Real sources is a critical step towards robust efficacy standards.

We invite researchers, institutions, and technology platforms to leverage this dataset in advancing the field of robust deepfake detection in voice. Together, we can develop next-generation safeguards against audio-based digital deception and preserve the integrity of our shared information ecosystem.