ID R&D scores lowest error rate for logical access in voice biometric spoofing challenge
The Global ASVspoof Challenge has been completed, with ID R&D announcing it achieved the highest accuracy evaluation in the logical access scenario.
The results for the 2019 Automatic Speaker Verification Spoofing and Countermeasures (ASVspoof) Challenge at Logical Access Condition (*1) show ID R&D, which is identified as T05, according to the announcement, ranked first with a Equal Error Rate (EER) of 0.22 percent. EER is the point at which false acceptances and false rejections are equal. Logical Access Condition refers to voice spoof attacks generated with advanced text-to-speech (TTS) synthesis and voice conversation (VC) technologies, according to the announcement. In a physical access scenario, T05 had the twelfth lowest EER among 52 spoof detection systems.
With fraudsters able to replicate a subject’s voice with less than a minute of voice samples, and increasing fears of deepfake capabilities, ID R&D cautions that spoofed speech could undermine business, personal, and government operations, making wide scale adoption of countermeasures urgent.
“As the number of crimes using synthetic or recorded voice increases, ID R&D is committed to delivering solutions that prevent even the most advanced spoofing attacks,” says ID R&D Chief Scientific Officer Konstantin Simonchik. “We are proud to have earned our top spot in the ASVspoof challenge and to contribute toward creating a future where interaction between users and technology is not just simple, but also secure.”
The ASVspoof challenge has been run as a biannual initiative since 2013 by concerned industry leaders, and ID R&D says it is the largest and most comprehensive evaluation of its kind. The 2019 challenge included participation from 63 research teams representing commercial and academic organizations.
ID R&D announced the appointment of a pair of industry veterans to its executive ranks earlier in September, and the company launched passive biometric liveness detection for both face and voice earlier in the year.