Saved in:
Bibliographic Details
Main Authors: Ibanez-Lissen, Luis, Gonzalez-Manzano, Lorena, de Fuentes, Jose Maria, Anciaux, Nicolas, Garcia-Alfaro, Joaquin
Format: Preprint
Published: 2024
Subjects:
Online Access:https://arxiv.org/abs/2411.19876
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866913643440898048
author Ibanez-Lissen, Luis
Gonzalez-Manzano, Lorena
de Fuentes, Jose Maria
Anciaux, Nicolas
Garcia-Alfaro, Joaquin
author_facet Ibanez-Lissen, Luis
Gonzalez-Manzano, Lorena
de Fuentes, Jose Maria
Anciaux, Nicolas
Garcia-Alfaro, Joaquin
contents Large Language Models (LLMs) are increasingly used in a variety of applications, but concerns around membership inference have grown in parallel. Previous efforts focus on black-to-grey-box models, thus neglecting the potential benefit from internal LLM information. To address this, we propose the use of Linear Probes (LPs) as a method to detect Membership Inference Attacks (MIAs) by examining internal activations of LLMs. Our approach, dubbed LUMIA, applies LPs layer-by-layer to get fine-grained data on the model inner workings. We test this method across several model architectures, sizes and datasets, including unimodal and multimodal tasks. In unimodal MIA, LUMIA achieves an average gain of 15.71 % in Area Under the Curve (AUC) over previous techniques. Remarkably, LUMIA reaches AUC>60% in 65.33% of cases -- an increment of 46.80% against the state of the art. Furthermore, our approach reveals key insights, such as the model layers where MIAs are most detectable. In multimodal models, LPs indicate that visual inputs can significantly contribute to detect MIAs -- AUC>60% is reached in 85.90% of experiments.
format Preprint
id arxiv_https___arxiv_org_abs_2411_19876
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle LUMIA: Linear probing for Unimodal and MultiModal Membership Inference Attacks leveraging internal LLM states
Ibanez-Lissen, Luis
Gonzalez-Manzano, Lorena
de Fuentes, Jose Maria
Anciaux, Nicolas
Garcia-Alfaro, Joaquin
Cryptography and Security
Artificial Intelligence
Large Language Models (LLMs) are increasingly used in a variety of applications, but concerns around membership inference have grown in parallel. Previous efforts focus on black-to-grey-box models, thus neglecting the potential benefit from internal LLM information. To address this, we propose the use of Linear Probes (LPs) as a method to detect Membership Inference Attacks (MIAs) by examining internal activations of LLMs. Our approach, dubbed LUMIA, applies LPs layer-by-layer to get fine-grained data on the model inner workings. We test this method across several model architectures, sizes and datasets, including unimodal and multimodal tasks. In unimodal MIA, LUMIA achieves an average gain of 15.71 % in Area Under the Curve (AUC) over previous techniques. Remarkably, LUMIA reaches AUC>60% in 65.33% of cases -- an increment of 46.80% against the state of the art. Furthermore, our approach reveals key insights, such as the model layers where MIAs are most detectable. In multimodal models, LPs indicate that visual inputs can significantly contribute to detect MIAs -- AUC>60% is reached in 85.90% of experiments.
title LUMIA: Linear probing for Unimodal and MultiModal Membership Inference Attacks leveraging internal LLM states
topic Cryptography and Security
Artificial Intelligence
url https://arxiv.org/abs/2411.19876