# Membership Inference Attack against Large Language Model-based Recommendation Systems: A New Distillation-based Paradigm

Source: [Membership Inference Attack against Large Language Model-based Recommendation Systems: A New Distillation-based Paradigm](<https://www.sota2.com/research/paper/691ea09cd6b0dd334292dfa4>)

Collection index: [llms.txt](<https://www.sota2.com/research/llms.txt>)

Data updated: 2026-03-21T05:00:04.296Z

Authors: Li Cuihong, Huang Xiaowen, Yin Chuanhuan, Sang Jitao

Published: 2025-09-16T09:36:43.000Z

## Abstract

Membership Inference Attack (MIA) aims to determine whether a specific data sample was included in the training dataset of a target model. Traditional MIA approaches rely on shadow models to mimic target model behavior, but their effectiveness diminishes for Large Language Model (LLM)-based recommendation systems due to the scale and complexity of training data. This paper introduces a novel knowledge distillation-based MIA paradigm tailored for LLM-based recommendation systems. Our method constructs a reference model via distillation, applying distinct strategies for member and non-member data to enhance discriminative capabilities. The paradigm extracts fused features (e.g., confidence, entropy, loss, and hidden layer vectors) from the reference model to train an attack model, overcoming limitations of individual features. Extensive experiments on extended datasets (Last.FM, MovieLens, Book-Crossing, Delicious) and diverse LLMs (T5, GPT-2, LLaMA3) demonstrate that our approach significantly outperforms shadow model-based MIAs and individual-feature baselines. The results show its practicality for privacy attacks in LLM-driven recommender systems.

## Recorded benchmark results

A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage.

- [Membership Inference Attack on Book-Crossing](<https://www.sota2.com/research/sota/membership-inference-attack-on-book-crossing.md>): Membership Inference Attack; Book-Crossing; Accuracy: 86.86.

- [Membership Inference Attack on Last.FM](<https://www.sota2.com/research/sota/membership-inference-attack-on-last-fm.md>): Membership Inference Attack; Last-FM; Accuracy: 65.09.

- [Membership Inference Attack on Delicious](<https://www.sota2.com/research/sota/membership-inference-attack-on-delicious.md>): Membership Inference Attack; Delicious; Accuracy: 66.36.

- [Membership Inference Attack on MovieLens](<https://www.sota2.com/research/sota/membership-inference-attack-on-movielens.md>): Membership Inference Attack; MovieLens; Accuracy: 85.

## Original sources and implementation

[Original paper](<https://arxiv.org/abs/2511.14763>)

[PDF](<https://arxiv.org/pdf/2511.14763.pdf>)
