EditMark: Watermarking Large Language Models based on Model Editing

2025-10-18

AI Summary

To address the challenge of embedding watermarks in large language models (LLMs) that simultaneously ensure copyright protection, zero performance degradation, and output naturalness, this paper proposes EditMark—the first lightweight watermarking method based on model editing. Unlike existing approaches requiring full retraining or fine-tuning, EditMark directly embeds verifiable watermarks into the model’s parameter space via adaptive multi-round weight editing and noise matrix injection—without any fine-tuning and without compromising generation quality. Its key contributions are: (1) the first application of model editing for LLM watermarking; (2) a unique watermark-response mechanism tailored for multi-answer tasks; and (3) efficient embedding of a 32-bit watermark in under 20 seconds—340× faster than fine-tuning—with 100% extraction accuracy, high fidelity, strong imperceptibility, and robustness against common adversarial attacks.

Technology Category

• Machine Learning: Large Multimodal Models (LMMs)
• Natural Language Processing: (Large) Language Models
• Computer Vision: Large Vision Models

Application Category

• Economics, Online Markets and Human Computation: LLM based quality controls for crowd work
• User Modeling, Personalization and Recommendation: Large Language Models (LLM) for user modeling and recommendation
• Search and Retrieval-Augmented AI: Search Tool Learning with LLM: Teaching LLMs to invoke search and make use of retrieved information

Abstract

Large Language Models (LLMs) have demonstrated remarkable capabilities, but their training requires extensive data and computational resources, rendering them valuable digital assets. Therefore, it is essential to watermark LLMs to protect their copyright and trace unauthorized use or resale. Existing methods for watermarking LLMs primarily rely on training LLMs with a watermarked dataset, which entails burdensome training costs and negatively impacts the LLM's performance. In addition, their watermarked texts are not logical or natural, thereby reducing the stealthiness of the watermark. To address these issues, we propose EditMark, the first watermarking method that leverages model editing to embed a training-free, stealthy, and performance-lossless watermark for LLMs. We observe that some questions have multiple correct answers. Therefore, we assign each answer a unique watermark and update the weights of LLMs to generate corresponding questions and answers through the model editing technique. In addition, we refine the model editing technique to align with the requirements of watermark embedding. Specifically, we introduce an adaptive multi-round stable editing strategy, coupled with the injection of a noise matrix, to improve both the effectiveness and robustness of the watermark embedding. Extensive experiments indicate that EditMark can embed 32-bit watermarks into LLMs within 20 seconds (Fine-tuning: 6875 seconds) with a watermark extraction success rate of 100%, which demonstrates its effectiveness and efficiency. External experiments further demonstrate that EditMark has fidelity, stealthiness, and a certain degree of robustness against common attacks.

Problem

• Protects LLM copyright without performance loss through model editing
• Creates stealthy watermarks using multiple correct answer variations
• Embeds robust watermarks rapidly without expensive retraining costs

Innovation

• Uses model editing for training-free watermark embedding
• Employs adaptive multi-round stable editing strategy
• Injects noise matrix to enhance watermark robustness