Multi bit watermarking protects generated audio under codec attacks
MARC: Multi-Bit Watermarking for Autoregressive Audio Generation against Codec Attacks
Cryptography and Security
Summary
It can be hard to tell if a piece of audio was created by a computer program, especially after the sound is compressed or changed. The researchers designed a new way, called MARC, to hide special watermarks in computer-generated audio that remain recognizable even after these changes. This method uses knowledge about how audio tokens behave and how codecs alter them to create watermarks that carry more than just a yes or no presence signal. MARC can recover detailed watermark information reliably, even after the audio goes through different compression processes or attempts to overwrite the watermark.
What this means in practice
- •For audio software developers: Embed robust multi-bit watermarks into generated speech or music that survive common audio compression for source verification.
- •For content distribution platforms: Identify and track generated audio content even after codec transformations to enforce content authenticity policies.$Commercial implications: Provides a way to sell watermarking-based content authentication and tracking services to streaming and social platforms.
Authors
Liaoran Xu, Weizhi Liu, Zhaoxia Yin
Abstract
Generated audio is now used in a range of applications, creating a need to verify its origin after distribution and signal processing. This task is particularly challenging for autoregressive audio generation because codec processing can alter the token sequence recovered from the waveform. Such changes reduce the reliability of watermark detection and payload decoding. Existing methods construct token groups using either intrinsic token representations or substitution patterns caused by transformations. As a result, intrinsic token relationships and codec induced substitutions are modeled separately. In addition, most methods support only zero bit detection. They can determine whether a watermark is present but cannot distinguish individual generated outputs. We propose \textbf{MARC}, a multi-bit generative watermarking method for autoregressive audio generation. MARC integrates intrinsic token representations with confusion patterns obtained through retokenization and multiple codecs, forming a codec-aware token-cluster space. Within this space, payload-driven cluster scheduling is used to embed a multi-bit watermark, while detection and payload decoding are performed on retokenized observations. Experiments on speech, dialogue, and music generation show that MARC achieves an average of 97.3\% bit extraction accuracy on unmodified watermarked audio and the watermark can still be extracted under diverse codec attacks. MARC also demonstrates robustness to overwriting attacks.