Multi-Target Backdoor Attacks Against Speaker Recognition

2025/08/12 by Fortier, Alexandrine, Sonal Joshi, Joshi, Sonal +8 · 1 citation
Computer Science · #Speech Recognition and Synthesis #Adversarial Robustness in Machine Learning #Speech and Audio Processing

paper · pdf · doi:10.48550/arxiv.2508.08559

Abstract

In this work, we propose a multi-target backdoor attack against speaker identification using position-independent clicking sounds as triggers. Unlike previous single-target approaches, our method targets up to 50 speakers simultaneously, achieving success rates of up to 95.04%. To simulate more realistic attack conditions, we vary the signal-to-noise ratio between speech and trigger, demonstrating a trade-off between stealth and effectiveness. We further extend the attack to the speaker verification task by selecting the most similar training speaker - based on cosine similarity - as a proxy target. The attack is most effective when target and enrolled speaker pairs are highly similar, reaching success rates of up to 90% in such cases.

Citations

• Breaking Speaker Recognition with PaddingBack
• Backdooring Instruction-Tuned Large Language Models with Virtual Prompt Injection
• Towards Stealthy Backdoor Attacks against Speech Recognition via Elements of Sound
• Instructions as Backdoors: Backdoor Vulnerabilities of Instruction Tuning for Large Language Models
• Going In Style: Audio Backdoors Through Stylistic Transformations
• Invisible Backdoor Attack with Sample-Specific Triggers
• Backdoor Attack against Speaker Verification
• VenoMave: Targeted Poisoning Against Speech Recognition
• Real-time, Universal, and Robust Adversarial Attacks Against Speaker Recognition Systems
• Who is Real Bob? Adversarial Attacks on Speaker Recognition Systems
• A backdoor attack against LSTM-based text classification systems
• BadNets: Identifying Vulnerabilities in the Machine Learning Model Supply Chain
• MUSAN: A Music, Speech, and Noise Corpus
• Poisoning Attacks against Support Vector Machines

Cited by

• Backdoor Attacks Against Speech Language Models

Related

• Backdoor Attack against Speaker Verification
• MASTERKEY: Practical Backdoor Attack Against Speaker Verification Systems
• Breaking Speaker Recognition with PaddingBack
• Dictionary Attacks on Speaker Verification
• Enrollment-stage Backdoor Attacks on Speaker Recognition Systems via Adversarial Ultrasound