Dictionary Attacks on Speaker Verification

TL;DR

This paper introduces a novel dictionary attack method against speaker verification systems, demonstrating that adversarially crafted voices can match a significant portion of a population, raising security concerns.

Contribution

The paper presents a generic formulation of dictionary attacks on speaker verification, using adversarial optimization to generate master voices effective across various systems and conditions.

Findings

Average 69% success rate for females

Average 38% success rate for males

Effective transferability between speaker encoders

Abstract

In this paper, we propose dictionary attacks against speaker verification - a novel attack vector that aims to match a large fraction of speaker population by chance. We introduce a generic formulation of the attack that can be used with various speech representations and threat models. The attacker uses adversarial optimization to maximize raw similarity of speaker embeddings between a seed speech sample and a proxy population. The resulting master voice successfully matches a non-trivial fraction of people in an unknown population. Adversarial waveforms obtained with our approach can match on average 69% of females and 38% of males enrolled in the target system at a strict decision threshold calibrated to yield false alarm rate of 1%. By using the attack with a black-box voice cloning system, we obtain master voices that are effective in the most challenging conditions and transferable between speaker encoders. We also show that, combined with multiple attempts, this attack opens even more to serious issues on the security of these systems.