Membership Inference Attacks on In-Context Learning Recommendation

ACL ARR 2026 January Submission6743 Authors

Keywords: In-Context Learning, Privacy and Security, Recommendation System

Abstract: Large language models (LLMs) based recommender systems (RecSys) can adapt flexibly across different domains. It uses in-context learning (ICL), i.e., prompts, including sensitive historical user-specific item interactions, to customize the recommendation functions. However, no study has examined whether such private information may be exposed by novel privacy attacks. We design several membership inference attacks (MIAs): \emph{Similarity, Memorization, Inquiry, and Poisoning attacks}, aiming to reveal whether system prompts include victims' historical interactions. We have carefully evaluated them on the latest open-source LLMs and three well-known RecSys datasets. The results confirm that the MIA threat to LLM RecSys is realistic, and that existing prompt-based defense methods may be insufficient to protect against these attacks.

Paper Type: Long

Research Area: Ethics, Bias, and Fairness

Research Area Keywords: In-Context Learning, Privacy and Security, Recommendation System

Contribution Types: Model analysis & interpretability, NLP engineering experiment, Publicly available software and/or pre-trained models, Data analysis

Languages Studied: English

Submission Number: 6743