WaterSeeker: Pioneering Efficient Detection of Watermarked Segments in Large Documents

TL;DR

WaterSeeker is a new method that efficiently detects and locates watermarked segments within large documents, balancing accuracy and computational efficiency, and enabling interpretable AI detection.

Contribution

It introduces a novel two-step approach combining anomaly detection and local verification for watermarked segment localization in large texts.

Findings

Achieves high detection accuracy with reduced computation time.

Effectively locates watermarked segments within extensive natural language documents.

Provides a foundation for interpretable AI detection systems.

Abstract

Watermarking algorithms for large language models (LLMs) have attained high accuracy in detecting LLM-generated text. However, existing methods primarily focus on distinguishing fully watermarked text from non-watermarked text, overlooking real-world scenarios where LLMs generate only small sections within large documents. In this scenario, balancing time complexity and detection performance poses significant challenges. This paper presents WaterSeeker, a novel approach to efficiently detect and locate watermarked segments amid extensive natural text. It first applies an efficient anomaly extraction method to preliminarily locate suspicious watermarked regions. Following this, it conducts a local traversal and performs full-text detection for more precise verification. Theoretical analysis and experimental results demonstrate that WaterSeeker achieves a superior balance between detection accuracy and computational efficiency. Moreover, its localization capability lays the foundation for building interpretable AI detection systems. Our code is available at https://github.com/THU-BPM/WaterSeeker.