Work

Tokens for Learning, Tokens for Unlearning: Mitigating Membership Inference Attacks in Large Language Models via Dual-Purpose Training

Year: 2025

Type: conference-paper

Abstract: Large language models (LLMs) have become the backbone of modern natural language processing but pose privacy concerns about leaking sensitive training data.Membership inference attacks (MIAs), which aim to infer whether a sample is included in a model's training dataset, can serve as a foundation for broader privacy threats.Existing defenses designed for traditional classification models do not account for the sequential nature of text data.As a result, they either require significant computational resources or fail to effectively mitigate privacy risks in LLMs.In this work, we propose DuoLearn, a lightweight yet effective empirical privacy defense for protecting training data of language models by leveraging token-specific characteristics.By analyzing token dynamics during training, we propose a token selection strategy that categorizes tokens into hard tokens for learning and memorized tokens for unlearning.Subsequently, our training-phase defense optimizes a novel dual-purpose tokenlevel loss to achieve a Pareto-optimal balance between utility and privacy.Extensive experiments demonstrate that our approach not only provides strong protection against MIAs but also improves language modeling performance by around 10% across various LLM architectures and datasets compared to the baselines.

DOI: 10.18653/v1/2025.findings-acl.1174

Authors: Toan Tran, Ruixuan Liu, Li Xiong

Institution: Emory University

Language: English

Topic: Privacy-Preserving Technologies in Data

Subfield: Artificial Intelligence

Field: Computer Science

Domain: Physical Sciences

Open Access status: gold

Funders: National Science Foundation, National Institutes of Health