Audio Model Watermarking: Methods, Metrics, and Unresolved Challenges

Abstract

The widespread adoption of audio models has brought their security vulnerabilities into sharp focus,particularly with regard to originality and intellectual property protection.Unauthorized replication and misuse pose significant threats to model owners,driving watermarking technology to emerge as a critical research frontier.This study offers a comprehensive analysis of key developments in audio model watermarking between 2020 and 2025,systematically classifying current methods,analyzing their benefits and limitations, identifying significant gaps in the current literature.The paper also offers a comprehensive examination of watermarking technologies for audio models and provides a set of metrics for performance evaluation.Finally,we conclude the paper with a discussion on future research directions in this field.

Keywords