Veridas upgrades voice biometrics engine for high-accuracy fraud detection and authentication
Veridas has launched a new voice biometrics engine which it says has dramatically improved its verification performance and spoof detection capabilities.
The company says the verification performance of its proprietary text independent and language independent solution improved by 40.39 percent, compared to its previous version.
The das-Peak biometric engine now has an equal error rate (EER) of 2.66 percent in the NIST SRE CTS20 challenge. Veridas’ algorithm has performed well in NIST testing so far.
Veridas says the accuracy of its biometric fraud detection has reached 98.62 percent in an internal evaluation based on 650,000 bona fide and spoof audio samples.
The company’s Liveness Detection solution also passed iBeta Quality Assurance Level 1 PAD (presentation attack detection) testing to the ISO/IEC 30107 standard in December. Veridas’ technology achieved a 0 percent attack presentation classification error rate (APCER) over 1,800 attacks attempted during the test.
The company raised $4.7 million in funding in late-2020, and also finished second in voice biometric accuracy among text-independent technologies in the worldwide SdSV (Short-duration Speaker Verification) Challenge in 2020.
Veridas’ biometric engine can authenticate users based on 3 seconds of audio in less than 140 milliseconds. The company also suggests its technology can be applied across call center environments, mobile applications, and third-party services like virtual assistants for omnichannel authentication.
Several companies are currently considering implementing Veridas’ voice biometrics, according to the announcement.
The post was updated at 3:17pm Eastern on February 5, 2021 to remove a specific reference to the company’s position on the challenge leaderboard at the request of NIST.
Article Topics
authentication | biometric liveness detection | biometric testing | biometrics | identity verification | NIST | spoof detection | Veridas | voice biometrics
Comments