Saved in:
Bibliographic Details
Main Authors: Mozaffari, Hamid, Marathe, Virendra J.
Format: Preprint
Published: 2024
Subjects:
Online Access:https://arxiv.org/abs/2406.10218
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866910487782883328
author Mozaffari, Hamid
Marathe, Virendra J.
author_facet Mozaffari, Hamid
Marathe, Virendra J.
contents Membership Inference Attacks (MIAs) determine whether a specific data point was included in the training set of a target model. In this paper, we introduce the Semantic Membership Inference Attack (SMIA), a novel approach that enhances MIA performance by leveraging the semantic content of inputs and their perturbations. SMIA trains a neural network to analyze the target model's behavior on perturbed inputs, effectively capturing variations in output probability distributions between members and non-members. We conduct comprehensive evaluations on the Pythia and GPT-Neo model families using the Wikipedia dataset. Our results show that SMIA significantly outperforms existing MIAs; for instance, SMIA achieves an AUC-ROC of 67.39% on Pythia-12B, compared to 58.90% by the second-best attack.
format Preprint
id arxiv_https___arxiv_org_abs_2406_10218
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Semantic Membership Inference Attack against Large Language Models
Mozaffari, Hamid
Marathe, Virendra J.
Machine Learning
Membership Inference Attacks (MIAs) determine whether a specific data point was included in the training set of a target model. In this paper, we introduce the Semantic Membership Inference Attack (SMIA), a novel approach that enhances MIA performance by leveraging the semantic content of inputs and their perturbations. SMIA trains a neural network to analyze the target model's behavior on perturbed inputs, effectively capturing variations in output probability distributions between members and non-members. We conduct comprehensive evaluations on the Pythia and GPT-Neo model families using the Wikipedia dataset. Our results show that SMIA significantly outperforms existing MIAs; for instance, SMIA achieves an AUC-ROC of 67.39% on Pythia-12B, compared to 58.90% by the second-best attack.
title Semantic Membership Inference Attack against Large Language Models
topic Machine Learning
url https://arxiv.org/abs/2406.10218