Whispeak Wins the Cyber Challenge for Deepfake Detection
Whispeak takes first place in the Cyber Challenge for detecting AI-generated speech!
The Defense Innovation Agency (AID), in collaboration with the Cyber Command (COMCYBER), recently organized the Cyber Challenge for detecting AI-generated speech, with technical evaluation conducted by AMIAD and the Directorate General of Armaments (DGA). This challenge brought together the best solutions to detect and analyze artificially generated voices, putting participants’ technologies to the test on a shared dataset.
Whispeak, a leader in voice biometrics and synthetic voice detection, is proud to announce its first-place finish in both categories of this demanding competition. Here’s a look back at this significant event and the innovative solutions that helped Whispeak stand out.
A Major Challenge in Audio Deepfake Detection
Organized by AID and COMCYBER, the Cyber Challenge for detecting AI-generated speech aimed to stimulate innovation in cybersecurity and address critical issues related to protecting voice-based systems. Two major technological categories were at the heart of this edition:
- Counter Measure & Automatic DeepFake Detection (ADD): This category focuses on detecting AI-generated speech without prior knowledge of the speaker, identifying any fraudulent voice modification or synthesis.
- Spoofing-Robust Automatic Speaker Verification (SASV): This category assesses the systems’ ability to detect audio deepfakes while targeting a specific speaker.
Participants had the opportunity to work in an “open conditions” environment, allowing them to use all available data to test and refine their models.
A Resounding Victory for Whispeak
Thanks to its expertise in voice-based applied research, Whispeak emerged as a leader in both challenge categories. This success underscores the excellence of our technologies in voice biometrics and synthetic voice detection, the two pillars of voice identity.
Our team of researchers and engineers developed solutions that combine precision, innovation, and performance to meet the high expectations of the challenge. The evaluations, based on metrics such as the Area Under the Receiver Operating Characteristic curve (AuROC), measured the systems’ ability to differentiate authentic signals from generated voices. Whispeak excelled in this test.
A Step Forward for Enhanced Cybersecurity
This victory is more than just recognition—it marks a major milestone for cybersecurity and voice authentication. The growing threats posed by audio deepfakes present a significant challenge for businesses and institutions. Whispeak is positioning itself as a key player in securing voice interactions against these risks.
Our technologies are designed to meet the most demanding environments and provide an essential layer of protection, particularly for strategic sectors such as defense, finance, and remote services.
We extend our thanks to AID, COMCYBER, AMIAD, DGA, and IRISA (Research Institute in Computer Science and Random Systems) for their commitment and vision in combating these emerging threats.
