mAVE: A Watermark for Joint Audio-Visual Generation Models

2026-03-07

AI Summary

This work addresses the vulnerability of existing watermarking methods in joint audio-visual generative models to Swap attacks, which can lead to forged content being falsely authenticated, thereby compromising copyright integrity and content trustworthiness. To counter this, the authors propose mAVE, a native watermarking framework specifically designed for joint generative models. By integrating inverse transform sampling with cryptographic binding during model initialization, mAVE constructs an entangled audio-visual manifold in the latent space, establishing strong cross-modal coupling. The method requires no fine-tuning and achieves watermark embedding without performance degradation on state-of-the-art models such as LTX-2 and MOVA. It ensures over 99% binding integrity and provides an exponential security margin, fundamentally mitigating audio-visual replacement attacks and safeguarding both content provenance and copyright.

Technology Category

• Computer Vision: Adversarial Attacks & Robustness
• Machine Learning: Large Multimodal Models (LMMs)
• Natural Language Processing: Language Grounding & Multi-modal NLP

Application Category

• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG
• User Modeling, Personalization and Recommendation: Attacks and countermeasures in recommendation systems
• Security and Privacy: Data transparency and provenance

Abstract

As Joint Audio-Visual Generation Models see widespread commercial deployment, embedding watermarks has become essential for protecting vendor copyright and ensuring content provenance. However, existing techniques suffer from an architectural mismatch by treating modalities as decoupled entities, exposing a critical Binding Vulnerability. Adversaries exploit this via Swap Attacks by replacing authentic audio with malicious deepfakes while retaining the watermarked video. Because current detectors rely on independent verification ($Video_{wm}\vee Audio_{wm}$), they incorrectly authenticate the manipulated content, falsely attributing harmful media to the original vendor and severely damaging their reputation. To address this, we propose mAVE (Manifold Audio-Visual Entanglement), the first watermarking framework natively designed for joint architectures. mAVE cryptographically binds audio and video latents at initialization without fine-tuning, defining a Legitimate Entanglement Manifold via Inverse Transform Sampling. Experiments on state-of-the-art models (LTX-2, MOVA) demonstrate that mAVE guarantees performance-losslessness and provides an exponential security bound against Swap Attacks. Achieving near-perfect binding integrity ($>99\%$), mAVE offers a robust cryptographic defense for vendor copyright.

Problem

• Joint Audio-Visual Generation
• Watermarking
• Binding Vulnerability
• Swap Attacks
• Content Provenance

Innovation

• audio-visual watermarking
• manifold entanglement
• swap attack resistance
• joint generative models
• cryptographic binding