The study investigates the use of neural replay simulators to improve the robustness of over-the-air adversarial attacks on Automatic Speaker Verification (ASV) systems. ASV systems, powered by deep learning, are known to be vulnerable to adversarial attacks, particularly in physical access scenarios. The researchers have conducted experiments on the ASVspoof2019 dataset and found that the neural replay simulator can significantly increase the success rates of over-the-air adversarial attacks. This raises concerns about the potential for adversarial attacks on speaker verification in physical access applications.
Publication date: 10 Oct 2023
Project Page: Not Provided
Paper: https://arxiv.org/pdf/2310.05354
