Automated Detection of Pre-training Text in Black-box LLMs

TL;DR

This paper introduces VeilProbe, an automated framework for detecting whether texts were part of an LLM's pre-training data in black-box scenarios, addressing privacy concerns without manual intervention.

Contribution

VeilProbe is the first automated, black-box capable method that infers membership of texts in LLM pre-training data without human-designed questions or instructions.

Findings

Effective detection on multiple datasets

Outperforms existing black-box methods

Reduces overfitting with prototype-based classifier

Abstract

Detecting whether a given text is a member of the pre-training data of Large Language Models (LLMs) is crucial for ensuring data privacy and copyright protection. Most existing methods rely on the LLM's hidden information (e.g., model parameters or token probabilities), making them ineffective in the black-box setting, where only input and output texts are accessible. Although some methods have been proposed for the black-box setting, they rely on massive manual efforts such as designing complicated questions or instructions. To address these issues, we propose VeilProbe, the first framework for automatically detecting LLMs' pre-training texts in a black-box setting without human intervention. VeilProbe utilizes a sequence-to-sequence mapping model to infer the latent mapping feature between the input text and the corresponding output suffix generated by the LLM. Then it performs the key token perturbations to obtain more distinguishable membership features. Additionally, considering real-world scenarios where the ground-truth training text samples are limited, a prototype-based membership classifier is introduced to alleviate the overfitting issue. Extensive evaluations on three widely used datasets demonstrate that our framework is effective and superior in the black-box setting.