Video Watermarking: Safeguarding Your Video from (Unauthorized) Annotations by Video-based LLMs

TL;DR

This paper presents Video Watermarking, a technique that embeds imperceptible watermarks into videos to prevent unauthorized annotations by video-based LLMs, thereby protecting video content and maintaining viewing quality.

Contribution

It introduces a novel multi-modal flow-based watermarking method that effectively safeguards videos against misuse by video-based LLMs, balancing imperceptibility and robustness.

Findings

Watermarking significantly reduces video comprehensibility by LLMs

Method maintains high visual quality of videos

Watermarks are robust against various attacks

Abstract

The advent of video-based Large Language Models (LLMs) has significantly enhanced video understanding. However, it has also raised some safety concerns regarding data protection, as videos can be more easily annotated, even without authorization. This paper introduces Video Watermarking, a novel technique to protect videos from unauthorized annotations by such video-based LLMs, especially concerning the video content and description, in response to specific queries. By imperceptibly embedding watermarks into key video frames with multi-modal flow-based losses, our method preserves the viewing experience while preventing misuse by video-based LLMs. Extensive experiments show that Video Watermarking significantly reduces the comprehensibility of videos with various video-based LLMs, demonstrating both stealth and robustness. In essence, our method provides a solution for securing video content, ensuring its integrity and confidentiality in the face of evolving video-based LLMs technologies.