A Statistical Hypothesis Testing Framework for Data Misappropriation Detection in Large Language Models

Fuente: arXiv
Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Cai, Yinpeng, Li, Lexin, Zhang, Linjun
Format: Preprint
Veröffentlicht: 2025
Schlagworte:
Online-Zugang:
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866912626669256704
author Cai, Yinpeng
Li, Lexin
Zhang, Linjun
author_facet Cai, Yinpeng
Li, Lexin
Zhang, Linjun
contents Large Language Models (LLMs) are rapidly gaining enormous popularity in recent years. However, the training of LLMs has raised significant privacy and legal concerns, particularly regarding the distillation and inclusion of copyrighted materials in their training data without proper attribution or licensing, an issue that falls under the broader concern of data misappropriation. In this article, we focus on a specific problem of data misappropriation detection, namely, to determine whether a given LLM has incorporated the data generated by another LLM. We propose embedding watermarks into the copyrighted training data and formulating the detection of data misappropriation as a hypothesis testing problem. We develop a general statistical testing framework, construct test statistics, determine optimal rejection thresholds, and explicitly control type I and type II errors. Furthermore, we establish the asymptotic optimality properties of the proposed tests, and demonstrate the empirical effectiveness through intensive numerical experiments.
format Preprint
id arxiv_https___arxiv_org_abs_2501_02441
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle A Statistical Hypothesis Testing Framework for Data Misappropriation Detection in Large Language Models
Cai, Yinpeng
Li, Lexin
Zhang, Linjun
Machine Learning
Artificial Intelligence
Computation and Language
Cryptography and Security
Statistics Theory
Large Language Models (LLMs) are rapidly gaining enormous popularity in recent years. However, the training of LLMs has raised significant privacy and legal concerns, particularly regarding the distillation and inclusion of copyrighted materials in their training data without proper attribution or licensing, an issue that falls under the broader concern of data misappropriation. In this article, we focus on a specific problem of data misappropriation detection, namely, to determine whether a given LLM has incorporated the data generated by another LLM. We propose embedding watermarks into the copyrighted training data and formulating the detection of data misappropriation as a hypothesis testing problem. We develop a general statistical testing framework, construct test statistics, determine optimal rejection thresholds, and explicitly control type I and type II errors. Furthermore, we establish the asymptotic optimality properties of the proposed tests, and demonstrate the empirical effectiveness through intensive numerical experiments.
title A Statistical Hypothesis Testing Framework for Data Misappropriation Detection in Large Language Models
topic Machine Learning
Artificial Intelligence
Computation and Language
Cryptography and Security
Statistics Theory
url https://arxiv.org/abs/2501.02441