Attacking Speaker Recognition With Deep Generative Models

• Main Authors: Cai, Wilson, Doshi, Anish, Valle, Rafael
• Format: Journal Article
• Language: English
• Published: 08-01-2018
• Subjects: Computer Science - Learning; Computer Science - Sound
• DOI: 10.48550/arxiv.1801.02384

In this paper we investigate the ability of generative adversarial networks
(GANs) to synthesize spoofing attacks on modern speaker recognition systems. We
first show that samples generated with SampleRNN and WaveNet are unable to fool
a CNN-based speaker recognition system. We propose a modification of the
Wasserstein GAN objective function to make use of data that is real but not
from the class being learned. Our semi-supervised learning method is able to
perform both targeted and untargeted attacks, raising questions related to
security in speaker authentication systems.