Novel Loss-Enhanced Universal Adversarial Patches for Sustainable Speaker Privacy

2025/05/26 by Elvir Karimov, А А Варламов, Karimov, Elvir +7 · 1 citation
Computer Science · #Adversarial Robustness in Machine Learning #Artificial Intelligence (cs.AI) #Audio and Speech Processing (eess.AS) #Cryptography and Security (cs.CR) #FOS: Computer and information sciences #FOS: Electrical engineering #Hate Speech and Cyberbullying Detection #Sound (cs.SD) #electronic engineering #information engineering

paper · pdf · doi:10.48550/arxiv.2505.19951

openalex publication_date 2025/05/26 · openalex created_date 2025/10/10 · openalex updated_date 2026/07/28

Abstract

Deep learning voice models are commonly used nowadays, but the safety processing of personal data, such as human identity and speech content, remains suspicious. To prevent malicious user identification, speaker anonymization methods were proposed. Current methods, particularly based on universal adversarial patch (UAP) applications, have drawbacks such as significant degradation of audio quality, decreased speech recognition quality, low transferability across different voice biometrics models, and performance dependence on the input audio length. To mitigate these drawbacks, in this work, we introduce and leverage the novel Exponential Total Variance (TV) loss function and provide experimental evidence that it positively affects UAP strength and imperceptibility. Moreover, we present a novel scalable UAP insertion procedure and demonstrate its uniformly high performance for various audio lengths.

Citations

• AASIST3: KAN-Enhanced AASIST Speech Deepfake Detection using SSL Features and Additional Regularization for the ASVspoof 2024 Challenge
• Starshaped compact hypersurfaces in warped product manifolds II: a class of Hessian type equations
• Certification of Speaker Recognition Models to Additive Perturbations
• Speaker anonymization using orthogonal Householder neural network
• CAM++: A Fast and Efficient Network for Speaker Verification Using Context-Aware Masking
• Robust Speech Recognition via Large-Scale Weak Supervision
• Wespeaker: A Research and Production oriented Speaker Embedding Learning Toolkit
• Analyzing Language-Independent Speaker Anonymization Framework under Unseen Conditions
• Adversarial Mask: Real-World Universal Adversarial Attack on Face Recognition Model
• SpeechBrain: A General-Purpose Speech Toolkit
• Conformer: Convolution-augmented Transformer for Speech Recognition
• Conformer: Convolution-augmented Transformer for Speech Recognition
• Enabling Fast and Universal Audio Adversarial Attack Using Generative Model
• Universal Adversarial Perturbations Generative Network for Speaker Recognition
• GeoDA: a geometric framework for black-box adversarial attacks
• Universal Adversarial Perturbations for Speech Recognition Systems
• RawNet: Advanced end-to-end deep neural network using raw waveforms for text-independent speaker verification
• Natural TTS Synthesis by Conditioning WaveNet on Mel Spectrogram Predictions
• Towards Deep Learning Models Resistant to Adversarial Attacks
• Universal adversarial perturbations
• Adam: A Method for Stochastic Optimization
• Explaining and Harnessing Adversarial Examples
• Intriguing properties of neural networks

Cited by

• Ghost poisoning: Making users invisible to speaker verification models

Related

• Whispering Under the Eaves: Protecting User Privacy Against Commercial and LLM-powered Automatic Speech Recognition Systems
• Improving Voice Quality in Speech Anonymization With Just Perception-Informed Losses
• Privacy-Utility Balanced Voice De-Identification Using Adversarial Examples
• VoxGuard: Evaluating User and Attribute Privacy in Speech via Membership Inference Attacks
• Anonymizing Speech with Generative Adversarial Networks to Preserve Speaker Privacy