mlsanityData to 14 Oct 2026

Revisiting the Robustness of Watermarking to Paraphrasing Attacks

arXiv 2411.05277 ↗First posted 8 Nov 2024cs.CR cs.CL cs.LGcited by 25

Not on the map. Why?

Cited by 22 in LLM safety and security, but placed by its own abstract, which didn’t fit.

Abstract (arXiv)

Amidst rising concerns about the internet being proliferated with content generated from language models (LMs), watermarking is seen as a principled way to certify whether text was generated from a model. Many recent watermarking techniques slightly modify the output probabilities of LMs to embed a signal in the generated output that can later be detected. Since early proposals for text watermarking, questions about their robustness to paraphrasing have been prominently discussed. Lately, some techniques are deliberately designed and claimed to be robust to paraphrasing. However, such watermarking schemes do not adequately account for the ease with which they can be reverse-engineered. We show that with access to only a limited number of generations from a black-box watermarked model, we can drastically increase the effectiveness of paraphrasing attacks to evade watermark detection, thereby rendering the watermark ineffective.

Key facts. Cited by 25 papers in the map (data through 14 Oct 2026). Builds on 14 of its 16 references that are in the map. First posted on arXiv 8 Nov 2024.

Builds on

14 of 16 references.

References · 14

  1. 2020Language Models are Few-Shot Learnerscited by 32,419
  2. 2019Exploring the Limits of Transfer Learning with a Unified Text-to-Text Transformercited by 14,599
  3. 2023Pythia: A Suite for Analyzing Large Language Models Across Training and Scalingcited by 1,786
  4. 2019GLTR: Statistical Detection and Visualization of Generated Textcited by 496
  5. 2018A Discourse-Aware Attention Model for Abstractive Summarization of Long Documentscited by 478
  6. 2023Paraphrasing evades detectors of AI-generated text, but retrieval is an effective defensecited by 429
  7. 2023Can AI-Generated Text be Reliably Detected?cited by 386
  8. 2023Robust Distortion-free Watermarks for Language Modelscited by 320
  9. 2023Provable Robust Watermarking for AI-Generated Textcited by 299
  10. 2021BookSum: A Collection of Datasets for Long-form Narrative Summarizationcited by 184
  11. 2020Pre-training via Paraphrasingcited by 92
  12. 2024Watermark Stealing in Large Language Modelscited by 92
2 more
  1. 2024An Entropy-based Text Watermarking Detection Methodcited by 75
  2. 2021Towards Document-Level Paraphrase Generation with Sentence Rewriting and Reorderingcited by 11

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2025: 8; 2026: 17.

25 papers.

Sort:

By area:

  1. 2025VLA-Mark: A cross modal watermark for large vision-language alignment modelcited by 12
  2. 2025RAG-WM: An Efficient Black-Box Watermarking Approach for Retrieval-Augmented Generation of Large Language Modelscited by 6
  3. 2025DITTO: A Spoofing Attack Framework on Watermarked LLMs via Knowledge Distillationcited by 6
  4. 2025Sandcastles in the Storm: Revisiting the (Im)possibility of Strong Watermarkingcited by 4
  5. 2026Watermarking LLM Agent Trajectoriescited by 4
  6. 2026Distilling the Thought, Watermarking the Answer: A Principle Semantic Guided Watermark for Large Reasoning Modelscited by 3
  7. 2026A Visual Semantic Adaptive Watermark grounded by Prefix-Tuning for Large Vision-Language Modelcited by 3
  8. 2025RLCracker: Evaluating the Worst-Case Vulnerability of LLM Watermarks with Adaptive RL Attackscited by 2
  9. 2025SimKey: A Semantically Aware Key Module for Watermarking Language Modelscited by 2
  10. 2025A Practical Synthesis of Detecting AI-Generated Textual, Visual, and Audio Contentcited by 1
  11. 2025MarkTune: Improving the Quality-Detectability Trade-off in Open-Weight LLM Watermarkingcited by 0
  12. 2026TimeMark: A Trustworthy Time Watermarking Framework for Exact Generation-Time Recovery from AIGCcited by 0
13 more
  1. 2026GoCoMA: Hyperbolic Multimodal Representation Fusion for Large Language Model-Generated Code Attributioncited by 0
  2. 2026LLM Output Detectability and Task Performance Can be Jointly Optimizedcited by 0
  3. 2026Linguistics-Aware Non-Distortionary LLM Watermarkingcited by 0
  4. 2026Beyond AI-Generated Labels: Watermarking, Co-Creation, and Conflation of AI-Generation with Disinformationcited by 0
  5. 2026DHMark: Public-Key Watermarking for LLM-Generated Text via Diffie-Hellman-Guided Rejection Samplingcited by 0
  6. 2026Stability-Aware Feature Design for Robust Watermark Detection in Machine-Generated Textcited by 0
  7. 2026OpenStamp: A Watermark for Open-Source Language Modelscited by 0
  8. 2026WoE Wrote It? Watermarking Mixture-of-Experts LLMs for Black-Box Text Provenancecited by 0
  9. 2026SemTrace: Source-Grounded Semantic Signatures for Tracing LLM Exposure to Protected Documentscited by 0
  10. 2026Flip, Don't Shuffle: Watermarking LLMs at the Speed of Inferencecited by 0
  11. 2026TrajMark: Ownership Attribution and Segment-Level Tamper Localization for Coding-Agent Trajectoriescited by 0
  12. 2026Invisible Ink, Visible Lies: How Production Watermarking Causes LLMs to Hallucinatecited by 0
  13. 2026Pin It on Anyone: Certified Targeted Displacement of Multi-Bit LLM Watermarks Without the Keycited by 0

Similar, not cited

Close in wording, no citation either way

  1. 2023On the Reliability of Watermarks for Large Language Models
  2. 2025Temperature Matters: Enhancing Watermark Robustness Against Paraphrasing Attacks
  3. 2025Signature vs. Substance: Evaluating the Balance of Adversarial Resistance and Linguistic Quality in Watermarking Large Language Models
  4. 2023On the Learnability of Watermarks for Language Models
  5. 2024Provably Robust Watermarks for Open-Source Language Models
  6. 2024Optimizing Adaptive Attacks against Watermarks for Language Models
  7. 2024Black-Box Detection of Language Model Watermarks
  8. 2024PostMark: A Robust Blackbox Watermark for Large Language Models