Certification of Speaker Recognition Models to Additive Perturbations
This work addresses robustness certification for voice biometrics systems, representing an incremental transfer of existing certification methods to a new domain.
The authors tackled the problem of certifying speaker recognition models against adversarial additive perturbations by transferring and improving randomized smoothing certification techniques from image to audio domain, demonstrating effectiveness on VoxCeleb datasets.
Speaker recognition technology is applied to various tasks, from personal virtual assistants to secure access systems. However, the robustness of these systems against adversarial attacks, particularly to additive perturbations, remains a significant challenge. In this paper, we pioneer applying robustness certification techniques to speaker recognition, initially developed for the image domain. Our work covers this gap by transferring and improving randomized smoothing certification techniques against norm-bounded additive perturbations for classification and few-shot learning tasks to speaker recognition. We demonstrate the effectiveness of these methods on VoxCeleb 1 and 2 datasets for several models. We expect this work to improve the robustness of voice biometrics and accelerate the research of certification methods in the audio domain.