Towards Optimal Statistical Watermarking
This provides a systematic statistical framework for watermarking AI-generated text, which is incremental but offers improved error rates and robustness for applications in content verification.
The authors tackled the problem of statistical watermarking by framing it as a hypothesis testing problem, establishing nearly matching upper and lower bounds on token requirements for small errors and achieving a rate of Θ(h⁻¹ log(1/h)) compared to previous h⁻² rates. They also characterized optimal robust watermarking via linear programming.
We study statistical watermarking by formulating it as a hypothesis testing problem, a general framework which subsumes all previous statistical watermarking methods. Key to our formulation is a coupling of the output tokens and the rejection region, realized by pseudo-random generators in practice, that allows non-trivial trade-offs between the Type I error and Type II error. We characterize the Uniformly Most Powerful (UMP) watermark in the general hypothesis testing setting and the minimax Type II error in the model-agnostic setting. In the common scenario where the output is a sequence of $n$ tokens, we establish nearly matching upper and lower bounds on the number of i.i.d. tokens required to guarantee small Type I and Type II errors. Our rate of $Θ(h^{-1} \log (1/h))$ with respect to the average entropy per token $h$ highlights potentials for improvement from the rate of $h^{-2}$ in the previous works. Moreover, we formulate the robust watermarking problem where the user is allowed to perform a class of perturbations on the generated texts, and characterize the optimal Type II error of robust UMP tests via a linear programming problem. To the best of our knowledge, this is the first systematic statistical treatment on the watermarking problem with near-optimal rates in the i.i.d. setting, which might be of interest for future works.