Anti-spoofing detection to prevent speech conversion, replay attacks, and TTS attacks.
Speech conversion, replay attacks, and TTS create certain signal artifacts that are sometimes indistinguishable by a human ear. ID R&D algorithms find and identify such artifacts to accurately determine liveness.
ID R&D invests heavily in voice anti-spoofing technology as we believe it is a critical component of strong authentication. Our IDLive Voice product leverages multiple techniques to passively determine liveness, including specific speech feature extraction, Gaussian Mixture Models, factor analysis, and deep convolutional neural networks that have been trained for spoofing detection.
ID R&D was the clear leader in the ASVspoof 2019: Automatic Speaker Verification Spoofing and Countermeasures Challenge for Logical Access (LA). The challenge is the largest of its kind and was designed to test the ability to distinguish between human voice and a synthesized human voice.
ID R&D was also a leader in the prior ASVspoof Challenge, ASVspoof 2017, which was designed to test biometric technology capabilities for “replay attack” detection (live voice vs a recording).