PostMark: A Robust Blackbox Watermark for Large Language Models

TL;DR

PostMark introduces a post-hoc, blackbox watermarking method for large language models that does not require access to model logits, enhancing robustness against paraphrasing and enabling third-party implementation.

Contribution

It presents PostMark, a novel post-hoc watermarking technique that is model-agnostic and more robust to paraphrasing attacks compared to existing methods.

Findings

PostMark is more robust to paraphrasing attacks.

It can be implemented without access to model logits.

The method maintains acceptable text quality.

Abstract

The most effective techniques to detect LLM-generated text rely on inserting a detectable signature -- or watermark -- during the model's decoding process. Most existing watermarking methods require access to the underlying LLM's logits, which LLM API providers are loath to share due to fears of model distillation. As such, these watermarks must be implemented independently by each LLM provider. In this paper, we develop PostMark, a modular post-hoc watermarking procedure in which an input-dependent set of words (determined via a semantic embedding) is inserted into the text after the decoding process has completed. Critically, PostMark does not require logit access, which means it can be implemented by a third party. We also show that PostMark is more robust to paraphrasing attacks than existing watermarking methods: our experiments cover eight baseline algorithms, five base LLMs, and three datasets. Finally, we evaluate the impact of PostMark on text quality using both automated and human assessments, highlighting the trade-off between quality and robustness to paraphrasing. We release our code, outputs, and annotations at https://github.com/lilakk/PostMark.