ID R&D has been granted a patent for integrated voice biometrics and spoof detection for keywords and short requests used to unlock and access mobile phones, smart and IoT devices to guard them against fraud carried out with generative AI.
The company notes in an announcement that the use of voice as an interface for smart and IoT devices is expected to rise at a 21.3 percent compound annual rate from last year through 2030. Advances in large language models (LLMs) is contributing to this growth, but the FTC notes it is also contributing to scams, according to the announcement.
The U.S. Patent and Trademark Office granted the patent for a “Method and Device for Unlocking a User Device by Voice.”
The patent document describes advanced neural networks matching the unique biometric characteristics of a user’s voice, as well as detecting whether the voice sample is real or fake, for what the patent refers to as “spoofing proof authentication.” The process can involve Fast Fourier Transform features and Discrete cosine transform features in the generation of three different confidence scores, which feed into an overall anti-spoofing confidence score.
“As a founder and biometrics veteran, I recognize that innovative technology is crucial for streamlining user experience and providing robust protection against fraud and unauthorized access, particularly in the voice domain,” says Alexey Khitrov, President of ID R&D. “By harnessing the inherent simplicity of voice control, we ensure secure access across a multitude of devices, including smartphones, PCs, and smart home systems. This positions ID R&D as a leader in the innovation of voice security.”
ID R&D has increasingly focused on detecting biometric presentation attacks, injection attacks and deepfakes, weighing in on a famous recent case of voice fraud and co-hosting an upcoming free webinar with Biometric Update on video injection.
Article: ID R&D patents voice biometrics and spoof detection to protect smart devices