# VoxWatermark: A Large-Scale Benchmark for Audio Watermark Detection under Perturbations

Source: [VoxWatermark: A Large-Scale Benchmark for Audio Watermark Detection under Perturbations](<https://www.sota2.com/research/paper/6a30d876d33dfb23e783cb59>)

Collection index: [llms.txt](<https://www.sota2.com/research/llms.txt>)

Data updated: 2026-06-17T05:00:05.807Z

Authors: Farnaz Sedaghati, Yuxi Wang, Zicheng Weng, Wei Rao

Published: 2026-06-13T08:28:06.000Z

## Abstract

With the rapid deployment of speech generation systems in open environments, providing verifiable source attribution and copyright accountability for audio content has become critical. A gap in current research is the lack of a unified benchmark that systematically compares different watermark injection methods under realistic distribution shifts. To address this, we build VoxWatermark by applying 10 watermarking methods (4 neural and 6 traditional) with unified injection and annotation on multilingual, multi-source corpora, and introducing no-box, black-box, and white-box perturbations to simulate real recording and transmission conditions. Based on this benchmark, we propose AudioWMD as a robust baseline detector for large-scale, multi-method, cross-distribution settings. Results show that injection-method diversity and distribution shifts affect detection stability, while validating the effectiveness and scalability of AudioWMD. Dataset and code are publicly available.

## Recorded benchmark results

A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage.

- [Audio Watermark Detection on VoxWatermark Set 1 (T1) (test)](<https://www.sota2.com/research/sota/audio-watermark-detection-on-voxwatermark-set-1-t1-test.md>): Audio Watermark Detection; VoxWatermark Set 1 (T1) (test); TPR: 98.

- [Audio Watermark Detection on VoxWatermark T2 (test)](<https://www.sota2.com/research/sota/audio-watermark-detection-on-voxwatermark-t2-test.md>): Audio Watermark Detection; VoxWatermark T2 (test); AUROC: 70.02.

- [Audio Watermark Detection on VoxWatermark (val)](<https://www.sota2.com/research/sota/audio-watermark-detection-on-voxwatermark-val.md>): Audio Watermark Detection; VoxWatermark (val); Overall AUROC: 88.3.

- [Audio Watermark Detection on VoxWatermark (test 2)](<https://www.sota2.com/research/sota/audio-watermark-detection-on-voxwatermark-test-2.md>): Audio Watermark Detection; VoxWatermark (test 2); Overall AUROC: 63.2.

- [Audio Watermark Detection on VoxWatermark 2 (test)](<https://www.sota2.com/research/sota/audio-watermark-detection-on-voxwatermark-2-test.md>): Audio Watermark Detection; VoxWatermark 2 (test); TPR: 77.78.

- [Watermark Detection on VoxPopuli-100k Scenario T1](<https://www.sota2.com/research/sota/watermark-detection-on-voxpopuli-100k-scenario-t1.md>): Watermark Detection; VoxPopuli-100k Scenario T1; AUC: 77.15.

## Original sources and implementation

[Original paper](<https://arxiv.org/abs/2606.15187>)

[PDF](<https://arxiv.org/pdf/2606.15187.pdf>)