Search papers, labs, and topics across Lattice.
AliMark tackles the vulnerability of sentence-level watermarking to structural paraphrasing attacks like sentence splitting and merging. It reframes watermarking as a bit sequence encoding and alignment problem, generating multiple restructured text variants and aligning their extracted bit sequences with the secret bit sequence to minimize alignment cost. Experiments show AliMark significantly outperforms existing methods under diverse paraphrasing attacks, demonstrating improved robustness.
Sentence-level watermarks can now survive aggressive paraphrasing attacks like sentence splitting and merging, thanks to a new alignment-based approach.
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.
