AliMark: Enhancing Robustness of Sentence-Level Watermarking Against Text Paraphrasing
Fuente:
arXiv
Gespeichert in:
| Hauptverfasser: | , , , , , , , |
|---|---|
| Format: | Preprint |
| Veröffentlicht: |
2026
|
| Schlagworte: | |
| Online-Zugang: | |
| Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
| _version_ | 1866916058804256768 |
|---|---|
| author | Li, Yuexin Qu, Wenjie Wu, Linyu Chen, Yulin He, Yufei Cao, Tri Hooi, Bryan Zhang, Jiaheng |
| author_facet | Li, Yuexin Qu, Wenjie Wu, Linyu Chen, Yulin He, Yufei Cao, Tri Hooi, Bryan Zhang, Jiaheng |
| contents | Existing sentence-level watermarking methods enhance robustness to paraphrasing by anchoring watermarks in sentence semantics. However, their prefix-based designs remain vulnerable to structural perturbations, such as sentence splitting and merging, which commonly arise under strong paraphrasers like DIPPER and GPT-3.5. To mitigate this issue, we propose AliMark, a framework that reformulates sentence-level watermarking as a bit sequence encoding and alignment problem between a potentially watermarked text and a secret bit sequence. Notably, our approach adopts a two-stage detection strategy: we generate multiple restructured text variants and adaptively align their extracted bit sequences with the secret bit sequence to minimize alignment cost. This multi-candidate alignment design naturally improves robustness to sentence merges and splits. Extensive experiments demonstrate that AliMark substantially outperforms state-of-the-art baselines under diverse paraphrasing attacks. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2605_29434 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | AliMark: Enhancing Robustness of Sentence-Level Watermarking Against Text Paraphrasing Li, Yuexin Qu, Wenjie Wu, Linyu Chen, Yulin He, Yufei Cao, Tri Hooi, Bryan Zhang, Jiaheng Cryptography and Security Artificial Intelligence Computation and Language Machine Learning Existing sentence-level watermarking methods enhance robustness to paraphrasing by anchoring watermarks in sentence semantics. However, their prefix-based designs remain vulnerable to structural perturbations, such as sentence splitting and merging, which commonly arise under strong paraphrasers like DIPPER and GPT-3.5. To mitigate this issue, we propose AliMark, a framework that reformulates sentence-level watermarking as a bit sequence encoding and alignment problem between a potentially watermarked text and a secret bit sequence. Notably, our approach adopts a two-stage detection strategy: we generate multiple restructured text variants and adaptively align their extracted bit sequences with the secret bit sequence to minimize alignment cost. This multi-candidate alignment design naturally improves robustness to sentence merges and splits. Extensive experiments demonstrate that AliMark substantially outperforms state-of-the-art baselines under diverse paraphrasing attacks. |
| title | AliMark: Enhancing Robustness of Sentence-Level Watermarking Against Text Paraphrasing |
| topic | Cryptography and Security Artificial Intelligence Computation and Language Machine Learning |
| url | https://arxiv.org/abs/2605.29434 |