E2E-VGuard: Adversarial Prevention for Production LLM-based End-To-End Speech Synthesis

Recent advancements in speech synthesis technology have enriched our daily lives, with high-quality and human-like audio widely adopted across real-world applications. However, malicious exploitation like voice-cloning fraud poses severe security risks. Existing defense techniques struggle to address the production large language model (LLM)-based speech synthesis. While previous studies have considered the protection for fine-tuning synthesizers, they assume manually annotated transcripts. Given the labor intensity of manual annotation, end-to-end (E2E) systems leveraging automatic speech recognition (ASR) to generate transcripts are becoming increasingly prevalent, e.g., voice cloning via commercial APIs. Therefore, this E2E speech synthesis also requires new security mechanisms. To tackle these challenges, we propose E2E-VGuard, a proactive defense framework for two emerging threats: (1) production LLM-based speech synthesis, and (2) the novel attack arising from ASR-driven E2E scenarios. Specifically, we employ the encoder ensemble with a feature extractor to protect timbre, while ASR-targeted adversarial examples disrupt pronunciation. Moreover, we incorporate the psychoacoustic model to ensure perturbative imperceptibility. For a comprehensive evaluation, we test 16 open-source synthesizers and 3 commercial APIs across Chinese and English datasets, confirming E2E-VGuard's effectiveness in timbre and pronunciation protection. Real-world deployment validation is also conducted. Our code and demo page are available at https://wxzyd123.github.io/e2e-vguard/.

• We introduce a more realistic and challenging scenario of end-to-end fine-tuning-based speech synthesis, and we propose a proactive framework, E2E-VGuard, to protect individual information.

• We consider defensive waveform disruption from the perspectives of timbre and pronunciation. For the timbre disruption, we propose a feature objective based on the encoder ensemble and feature extractor. For the pronunciation disruption, we utilize AEs against ASR systems to fool TTS models with incorrect text and impact pronunciation.

• We utilize the psychoacoustic model with ℓ 2 -norm to enhance the perturbation imperceptibility for better human audible perception.

• We evaluate the effectiveness, transferability, and robustness of E2E-VGuard through comprehensive experiments across diverse settings: 19 TTS models (including 16 open-source and 3 commercial), 7 ASR systems, and 3 English and Chinese datasets.