Daon is integrating a patented synthetic voice detection technology into its call center fraud protection platform. This technology enhances voice biometric authentication by analyzing voice data through stereophonic processing. The method aims to detect subtle anomalies associated with deepfakes, thus improving security measures. Additional applications are planned across various industries, emphasizing the importance of safeguarding against advanced voice spoofing techniques.
Daon is set to integrate a newly patented synthetic voice detection technology into its call center fraud protection platform. This patent, titled “Methods And Systems For Enhancing The Detection Of Synthetic Voice Data,” received approval from the U.S. Patent and Trademark Office. It outlines a process where monophonic voice data is transformed into stereophonic data, which is then analyzed using a machine learning model to detect synthetic artifacts.
Ralph Rodriguez, Daon’s President and CPO, is a co-inventor of the patent alongside Olena and Davyd Mizynchuk. Daon’s research team, comprised of approximately 50 specialists, focuses on innovations related to biometrics, AI, and deepfake detection. This patented technology will enhance the capabilities of Daon’s xDeTECH software, aimed at detecting altered voices and ensuring security in voice biometric authentication.
Rodriguez highlights the vulnerability of voice authentication systems without protective measures against sophisticated spoofing techniques. He points out that the absence of synthetic voice detection could allow deepfake voices generated by tools like ElevenLabs to succeed in biometric voice authentication. This technology aims to identify anomalies unique to synthetic voices, thereby offering an essential protective layer against typical biometric vulnerabilities.
The patent’s technology can also integrate with leading telephony systems such as Genesys, Amazon Connect, Cisco, and Avaya. Additional applications could include securing IT help desk operations and protecting systems in telecom, healthcare, and government sectors. Rodriguez emphasizes that many deepfake generation methods fail to accurately reproduce the complexities of natural speech, which can result in detectable anomalies.
To combat this, the patented method involves processing audio data in stereo, facilitating thorough analysis by identifying subtle frequency anomalies, phase shifts, and harmonic structure inconsistencies not visible in monophonic audio data. Rodriguez has also received notification from the USPTO regarding approval of his upcoming patent for “Methods and Systems for Enhancing Detection of Fraudulent Data” expected by year-end.
Daon’s patent on synthetic voice detection technology represents a significant advancement in securing voice biometric authentication. By utilizing stereo processing and machine learning, the technology aims to effectively identify synthetic and manipulated voices, enhancing fraud protection in call centers and other sectors. The integration of such innovations could mitigate vulnerabilities in existing voice authentication systems, offering a multi-layered security approach against potential deepfake threats.
Original Source: www.biometricupdate.com
